rustc: Tweak `#[target_feature]` syntax

This is an implementation of the `#[target_feature]` syntax-related changes of
[RFC 2045][rfc]. Notably two changes have been implemented:

* The new syntax is `#[target_feature(enable = "..")]` instead of
  `#[target_feature = "+.."]`. The `enable` key is necessary instead of the `+`
  to indicate that a feature is being enabled, and a sub-list is used for
  possible expansion in the future. Additionally within this syntax the feature
  names being enabled are now whitelisted against a known set of target feature
  names that we know about.

* The `#[target_feature]` attribute can only be applied to unsafe functions. It
  was decided in the RFC that invoking an instruction possibly not defined for
  the current processor is undefined behavior, so to enable this feature for now
  it requires an `unsafe` intervention.

[rfc]: https://github.com/rust-lang/rfcs/blob/master/text/2045-target-feature.md

BufRead: Only flush the internal buffer if seeking outside of it.

Fixes #31100

r? @dtolnay

This'll enable running queries that could be cached and overall be more amenable
to the query infastructure.

Remove `impl Foo for .. {}` in favor `auto trait Foo {}`

Rebase of https://github.com/rust-lang/rust/pull/46480 with restored parsing support.

This is an implementation of the `#[target_feature]` syntax-related changes of
[RFC 2045][rfc]. Notably two changes have been implemented:

* The new syntax is `#[target_feature(enable = "..")]` instead of
  `#[target_feature = "+.."]`. The `enable` key is necessary instead of the `+`
  to indicate that a feature is being enabled, and a sub-list is used for
  possible expansion in the future. Additionally within this syntax the feature
  names being enabled are now whitelisted against a known set of target feature
  names that we know about.

* The `#[target_feature]` attribute can only be applied to unsafe functions. It
  was decided in the RFC that invoking an instruction possibly not defined for
  the current processor is undefined behavior, so to enable this feature for now
  it requires an `unsafe` intervention.

[rfc]: https://github.com/rust-lang/rfcs/blob/master/text/2045-target-feature.md

Also refactored parsing auto traits.

`fn check_impl` was feeling lonely with a file all for itself.

The WF checks are now done as an AST validation.

No longer parse it.
Remove AutoTrait variant from AST and HIR.
Remove backwards compatibility lint.
Remove coherence checks, they make no sense for the new syntax.
Remove from rustdoc.

Remove deprecated unstable attribute #[simd]

The `#[simd]` attribute has been deprecated since c8b6d5b2 back in 2015. Any nightly crates using it have had ample time to switch to `#[repr(simd)]`, and if they didn't they're likely broken by now anyway.

r? @EddyB

`struct` pattern parsing and diagnostic tweaks

 - Recover from struct parse error on match and point out missing match
   body.
 - Point at struct when finding non-identifier while parsing its fields.
 - Add label to "expected identifier, found {}" error.

Fix #15980.

Use DefIndex encoding that works better with on-disk variable length integer representations.

Use the least instead of the most significant bit for representing the address space.

r? @EddyB

Fix 45345

There is a fix for https://github.com/rust-lang/rust/issues/45345

It re-introduces `CFG_LIBDIR_RELATIVE` which was broken when migration from `configure` script to `x.py`.

Other commits fix errors which happen after rustbuild cleanups.

 type error method suggestions use whitelisted identity-like conversions

![method_jamboree_summit](https://user-images.githubusercontent.com/1076988/33523646-e5c43184-d7c0-11e7-98e5-1bff426ade86.png)

Previously, on a type mismatch (and if this wasn't preëmpted by a
higher-priority suggestion), we would look for argumentless methods
returning the expected type, and list them in a `help` note. This had two
major shortcomings: firstly, a lot of the suggestions didn't really make
sense (if you used a &str where a String was expected,
`.to_ascii_uppercase()` is probably not the solution you were hoping
for). Secondly, we weren't generating suggestions from the most useful
traits! We address the first problem with an internal
`#[rustc_conversion_suggestion]` attribute meant to mark methods that keep
the "same value" in the relevant sense, just converting the type. We
address the second problem by making `FnCtxt.probe_for_return_type` pass
the `ProbeScope::AllTraits` to `probe_op`: this would seem to be safe
because grep reveals no other callers of `probe_for_return_type`.

Also, structured suggestions are pretty and good for RLS and friends.

Unfortunately, the trait probing is still not all one would hope for: at a
minimum, we don't know how to rule out `into()` in cases where it wouldn't
actually work, and we don't know how to rule in `.to_owned()` where it
would. Issues #46459 and #46460 have been filed and are ref'd in a FIXME.

This is hoped to resolve #42929, #44672, and #45777.

Rollup of 24 pull requests

- Successful merges: #46985, #47069, #47081, #47185, #47282, #47283, #47288, #47289, #47298, #47305, #47306, #47307, #47310, #47324, #47328, #47331, #47340, #47343, #47344, #47352, #47357, #47365, #47375, #47382
- Failed merges: #47334

Ignore CTFE errors while lowering patterns

Closes #43105.

r? @EddyB

Fix examples of Duration::subsec_millis and Duration::subsec_micros

Update examples of `Duration::subsec_millis` and `Duration::subsec_micros`, because they are not for these two methods actually.

Re-enable num tests on wasm

Issue #42630 was closed but the tests are still ignored, supposedly they should pass now.

Fix docs for OsStr

At present, there are two small issues with the [docs](https://doc.rust-lang.org/std/ffi/struct.OsStr.html) for std::ffi::OsStr:
- The docs say "OsStr is to OsString as String is to &str: the former in each pair are borrowed references; the latter are owned strings.". The latter pair is mixed up: String is the owned variant whereas &str is the borrowed reference.
- The doc links to String and &str are broken and render as [String] and [&str].

This PR fixes these issues.

Make libtest build on CloudABI.

Just like on UNIX systems, we need to use sysconf() to obtain the number
of CPUs. Extend the existing cfg()'s to match CloudABI as well.

Add tests to fixed issues.

Closes #36792. Closes #38091. Closes #39687. Closes #42148. Closes #42956.

Glued tokens can themselves be joint.

When gluing two tokens, the second of which is joint, the result should also be
joint.
This fixes an issue with joining three `Dot` tokens to make a `DotDotDot` - the
intermediate `DotDot` would not be joint and therefore we would not attempt to
glue the last `Dot` token, yielding `.. .` instead of `...`.

r? @jseyfried

Fix typo.

Add -Ztime-passes line for dep-graph loading.

We measure how much time the background thread spends on loading the dep-graph but not how long the main thread is blocked while the background thread is still working. Let's change that!

Use the new fs_read_write functions in rustc internals

Uses `fs::read` and `fs::write` (added by #45837) where appropriate, to simplify code and dog-food these new APIs.  This also improves performance, when combined with #47324.

Pre-allocate in fs::read and fs::read_string

This is a simpler alternative to #46340 and #45928, as requested by the libs team.

fix typo rwlock.rs

Hi. Fixed typo: contained -> content

Fix panic strings.

 - Fix panic string in `check_ast_crate`.
 - Update panic string for Duration subtraction on overflow/underflow.

Not sure if the changes to `Duration` are helpful/needed. Mostly just a nit.
Otherwise this is just a one character change :)

On another note: I hit the panic in `check_ast_crate` when compiling
[m-labs/smoltcp] with the following:

```
cargo test --doc --no-default-features --features "std socket-raw"`
```

[m-labs/smoltcp]: https://github.com/m-labs/smoltcp

Don't track local_needs_drop separately in qualify_consts.

None