Parse union items and add a feature for them

rustc: Implement custom derive (macros 1.1)

This commit is an implementation of [RFC 1681] which adds support to the
compiler for first-class user-define custom `#[derive]` modes with a far more
stable API than plugins have today.

[RFC 1681]: https://github.com/rust-lang/rfcs/blob/master/text/1681-macros-1.1.md

The main features added by this commit are:

* A new `rustc-macro` crate-type. This crate type represents one which will
  provide custom `derive` implementations and perhaps eventually flower into the
  implementation of macros 2.0 as well.

* A new `rustc_macro` crate in the standard distribution. This crate will
  provide the runtime interface between macro crates and the compiler. The API
  here is particularly conservative right now but has quite a bit of room to
  expand into any manner of APIs required by macro authors.

* The ability to load new derive modes through the `#[macro_use]` annotations on
  other crates.

All support added here is gated behind the `rustc_macro` feature gate, both for
the library support (the `rustc_macro` crate) as well as the language features.

There are a few minor differences from the implementation outlined in the RFC,
such as the `rustc_macro` crate being available as a dylib and all symbols are
`dlsym`'d directly instead of having a shim compiled. These should only affect
the implementation, however, not the public interface.

This commit also ended up touching a lot of code related to `#[derive]`, making
a few notable changes:

* Recognized derive attributes are no longer desugared to `derive_Foo`. Wasn't
  sure how to keep this behavior and *not* expose it to custom derive.

* Derive attributes no longer have access to unstable features by default, they
  have to opt in on a granular level.

* The `derive(Copy,Clone)` optimization is now done through another "obscure
  attribute" which is just intended to ferry along in the compiler that such an
  optimization is possible. The `derive(PartialEq,Eq)` optimization was also
  updated to do something similar.

---

One part of this PR which needs to be improved before stabilizing are the errors
and exact interfaces here. The error messages are relatively poor quality and
there are surprising spects of this such as `#[derive(PartialEq, Eq, MyTrait)]`
not working by default. The custom attributes added by the compiler end up
becoming unstable again when going through a custom impl.

Hopefully though this is enough to start allowing experimentation on crates.io!

Introduce max_by/min_by on iterators

See https://github.com/rust-lang/rfcs/issues/1722 for reference.

It seems that there is `min`, `max` (simple computation of min/max), `min_by_key`, `max_by_key` (min/max by comparing mapped values) but no `min_by` and `max_by` (min/max according to comparison function). However, e.g. on vectors or slices there is `sort`, `sort_by_key` and `sort_by`.

Rollup of 12 pull requests

- Successful merges: #35754, #35793, #36099, #36160, #36171, #36178, #36180, #36190, #36198, #36205, #36210, #36223
- Failed merges:

Update compiler error E0558 to use new error format

Fixes #36196 part of #35233

r? @jonathandturner

Fixed E0529's label and unit test

Fixes #36195 part of #35233.

This is ready for review, but will likely fail Travis due to #36138. I changed the wording of the label, so feedback on that would be appreciated.

r? @jonathandturner

Fixed E0528 label and unit test

Fixes #36194 part of #35233.

r? @jonathandturner

test: Add a min-llvm-version directive

We've got tests which require a particular version of LLVM to run as they're
testing bug fixes. Our build system, however, supports multiple LLVM versions,
so we can't run these tests on all LLVM versions.

This adds a new `min-llvm-version` directive for tests so they can opt out of
being run on older versions of LLVM. This then namely applies that logic to the
`issue-36023.rs` test case and...

Closes #36138

configure: check if any of the arguments contain --help

Currently it checks only the first argument.

Fixes #31216

Transition Travis CI to use rustbuild.

Special case a few colors for Windows

As brought up on [this thread](https://github.com/rust-lang/rust/issues/33240#issuecomment-243310067) the colors used in error messages on Windows can be difficult to read because of the lack of bold.

This PR makes a few changes to improve readability, namely:
* Rather than using BRIGHT_BLUE, on Windows we now use BRIGHT_CYAN, which is easier to read on black when you do not have bold
* We used BRIGHT_YELLOW rather than YELLOW, for the same reason
* Titles will be BRIGHT_WHITE now, to give the illusion of being bold

Some examples:

![warning](https://cloud.githubusercontent.com/assets/547158/18148466/9aa9bbe2-6f8e-11e6-927f-d0eec53cac32.PNG)

![error](https://cloud.githubusercontent.com/assets/547158/18148488/ba9fb186-6f8e-11e6-8d8e-e93d569f61de.PNG)

r? @nikomatsakis

cc @retep998

Update lifetime errors to specifically note temporaries

This PR updates the error message we give in the case of a temporary value not living long enough.

Before:

<img width="497" alt="screen shot 2016-08-31 at 10 02 47 am" src="https://cloud.githubusercontent.com/assets/547158/18138551/27a06794-6f62-11e6-9ee2-bdf8bed75ca7.png">

Now:

<img width="488" alt="screen shot 2016-08-31 at 10 03 01 am" src="https://cloud.githubusercontent.com/assets/547158/18138557/2e5cf322-6f62-11e6-9047-4a78abf3d78c.png">

Specifically, it makes the following changes:

* Detects if a temporary is being used.  If so, it changes the labels to mention that a temporary value specifically is in question
* Simplifies wording of the existing labels to focus on lifetimes rather than values being valid
* Changes the help to a note, since the help+span wasn't as helpful (and sometimes more confusing) than just a note.

r? @nikomatsakis

Normalize the function signature of closures

Previously we didn't normalize the function signatures used for
closures. This didn't cause a problem in most cases, but caused an ICE
in during MIR type checking.

Fixes #36139

r? @EddyB

Document try!'s error conversion behaviour

try!'s documentation currently doesn't document the error conversion behaviour of the macro. This patch extends the documentation.

Open questions:
* is it worthwhile to have seperate examples with and without wrapping behaviour? It's not immediately obvious that From<T> for T is always defined. Though this is necessary for the macro to work in any case, is this the place to expect that knowledge.

demonstrate `RHS != Self` use cases for `Add` and `Sub`

Add `must_use` to the Reference

I'm a bit uncertain about the exact phrasing, but having it mentioned at all is probably better than before.

Fixes #36196 part of #35233

This commit is an implementation of [RFC 1681] which adds support to the
compiler for first-class user-define custom `#[derive]` modes with a far more
stable API than plugins have today.

[RFC 1681]: https://github.com/rust-lang/rfcs/blob/master/text/1681-macros-1.1.md

The main features added by this commit are:

* A new `rustc-macro` crate-type. This crate type represents one which will
  provide custom `derive` implementations and perhaps eventually flower into the
  implementation of macros 2.0 as well.

* A new `rustc_macro` crate in the standard distribution. This crate will
  provide the runtime interface between macro crates and the compiler. The API
  here is particularly conservative right now but has quite a bit of room to
  expand into any manner of APIs required by macro authors.

* The ability to load new derive modes through the `#[macro_use]` annotations on
  other crates.

All support added here is gated behind the `rustc_macro` feature gate, both for
the library support (the `rustc_macro` crate) as well as the language features.

There are a few minor differences from the implementation outlined in the RFC,
such as the `rustc_macro` crate being available as a dylib and all symbols are
`dlsym`'d directly instead of having a shim compiled. These should only affect
the implementation, however, not the public interface.

This commit also ended up touching a lot of code related to `#[derive]`, making
a few notable changes:

* Recognized derive attributes are no longer desugared to `derive_Foo`. Wasn't
  sure how to keep this behavior and *not* expose it to custom derive.

* Derive attributes no longer have access to unstable features by default, they
  have to opt in on a granular level.

* The `derive(Copy,Clone)` optimization is now done through another "obscure
  attribute" which is just intended to ferry along in the compiler that such an
  optimization is possible. The `derive(PartialEq,Eq)` optimization was also
  updated to do something similar.

---

One part of this PR which needs to be improved before stabilizing are the errors
and exact interfaces here. The error messages are relatively poor quality and
there are surprising spects of this such as `#[derive(PartialEq, Eq, MyTrait)]`
not working by default. The custom attributes added by the compiler end up
becoming unstable again when going through a custom impl.

Hopefully though this is enough to start allowing experimentation on crates.io!

syntax-[breaking-change]

Address FIXME in libcollectionstest/btree/set.rs

add mips64-gnu and mips64el-gnu targets

With this commit one can build no_core (and probably no_std as well)
Rust programs for these targets. It's not yet possible to cross compile
std for these targets because rust-lang/libc doesn't know about the
mips64 architecture.

These targets have been tested by cross compiling the "smallest hello"
program (see code below) and then running it under QEMU.

``` rust

extern {
    fn puts(_: *const u8);
}

fn start(_: isize, _: *const *const u8) -> isize {
    unsafe {
        let msg = b"Hello, world!\0";
        puts(msg as *const _ as *const u8);
    }
    0
}

trait Copy {}

trait Sized {}
```

cc #36015
r? @alexcrichton
cc @brson

The cabi stuff is likely wrong. I just copied cabi_mips source and changed some `4`s to `8`s and `32`s to `64`s. It was enough to get libc's `puts` to work but I'd like someone familiar with this module to check it.

implementing RFC 1623. This fixes #35897.

This is a work in progress. In particular, I want to add more tests,
especially the compile-fail test is very bare-bones.

Implement RFC 1560 behind `#![feature(item_like_imports)]`

This implements https://github.com/rust-lang/rfcs/pull/1560 (cc #35120) behind the `item_like_imports` feature gate.

The [RFC text](https://github.com/rust-lang/rfcs/blob/master/text/1560-name-resolution.md#changes-to-name-resolution-rules) describes the changes to name resolution enabled by `#![feature(item_like_imports)` in detail. To summarize,
 - Items and named imports shadow glob imports.
 - Multiple globs can import the same name if the name is unused or the imports are shadowed.
 - Multiple globs can import the same name if the imports are of the same item (following re-exports).
  - The visibility of such a name is the maximum visibility of the imports.
  - Equivalently, adding a glob import will never reduce the visibility of a name, nor will removing one increase it.
 - Non-prelude private imports can be used wherever we currently allow private items to be used.
  - Prelude-imported names are unaffected, i.e. they continue to be usable only in lexical scopes.
 - Globs import all visible names, not just public names.
  - Equivalently, glob importing from an ancestor module imports all of the ancestor's names, and glob importing from other modules is unchanged.

r? @nrc

We've got tests which require a particular version of LLVM to run as they're
testing bug fixes. Our build system, however, supports multiple LLVM versions,
so we can't run these tests on all LLVM versions.

This adds a new `min-llvm-version` directive for tests so they can opt out of
being run on older versions of LLVM. This then namely applies that logic to the
`issue-36023.rs` test case and...

Closes #36138

to reexport some (but not all) names with less visibility.

to reexport some (but not all) namespaces with less visibility.