rustbuild: Implement `make dist`

This commit implements the `make dist` command in the new rustbuild build
system, porting over `dist.mk` and `prepare.mk` into Rust. There's a huge amount
of complexity between those two files, not all of which is likely justified, so
the Rust implementation is *much* smaller.

Currently the implementation still shells out to rust-installer as well as some
python scripts, but ideally we'd rewrite it all in the future to not shell out
and be in Rust proper.

const eval: don't assume the rhs of a bitshift is of any particular type

[regression found](https://internals.rust-lang.org/t/regression-report-stable-2016-03-03-vs-nightly-2016-03-15/3250) in jpeg-decoder

This commit implements the `make dist` command in the new rustbuild build
system, porting over `dist.mk` and `prepare.mk` into Rust. There's a huge amount
of complexity between those two files, not all of which is likely justified, so
the Rust implementation is *much* smaller.

Currently the implementation still shells out to rust-installer as well as some
python scripts, but ideally we'd rewrite it all in the future to not shell out
and be in Rust proper.

Resolve: improve diagnostics for duplicate definitions and imports

This PR improves and regularizes the diagnostics for duplicate definitions and imports.

After this PR, the second of two duplicate definitions/imports will have the following form:
> a(n) [value|type|module|trait|extern crate] named \`*name*\` has already been [defined|imported] in this [module|block|trait|enum]

with a note referencing this first of the two duplicate definitions/imports:
> previous [definition|import] of \`*name*\` here

The error indices remain unchanged.

r? @EddyB

syntax: Always pretty print a newline after doc comments

Before this patch, code that had a doc comment as the first
line, as in:

```rust
/// Foo
struct Foo;
```

Was pretty printed into:

```rust
///Foostruct Foo;
```

This makes sure that that there is always a trailing newline
after a doc comment.

Closes #31722

Normalize return type when checking for E0269

Fixes #31597

First time dealing with normalization. Maybe `normalize_associated_type` would be better here, but it seems to imply it's only used during trans.

update compiler-rt submodule

fixes #32194

also see rust-lang/compiler-rt#17

r? @alexcrichton

Fixes #31597

std: Fix overflow when subtracting Instant

This code was currently only exercised on OSX, but this applies the same method
of subtraction used on Linux which doesn't have the same overflow issues.

Note that this currently includes no tests, but that's because this is only
visible with debug assertions enabled. Soon, however, I'll enable debug
assertions on all auto builds on the bots so we should get testing for this.

Closes #32268

rustbuild: Fix LLVM compile on MinGW

The LLVM change [1] in #32239 unfortunately broke the LLVM build on MinGW, so
this LLVM submodule update brings in one more fix [2] which should hopefully
remedy that.

Once this lands we should be able to immediately start gating on this to prevent
it from happening again.

[1]: https://github.com/rust-lang/llvm/commit/be89e4b5
[2]: https://github.com/rust-lang/llvm/commit/3dcd2c84

fixes #32194

rustc: Improve compile time of platform intrinsics

This commit improves the compile time of `rustc_platform_intrinsics` from 23s to
3.6s if compiling with `-O` and from 77s to 17s if compiling with `-O -g`. The
compiled rlib size also drops from 3.1M to 1.2M.

The wins here were gained by removing the destructors associated with `Type` by
removing the internal `Box` and `Vec` indirections. These destructors meant that
a lot of landing pads and extra code were generated to manage the runtime
representations. Instead everything can basically be statically computed and
shoved into rodata, so all we need is a giant string compare to lookup what's
what.

Closes #28273

This commit improves the compile time of `rustc_platform_intrinsics` from 23s to
3.6s if compiling with `-O` and from 77s to 17s if compiling with `-O -g`. The
compiled rlib size also drops from 3.1M to 1.2M.

The wins here were gained by removing the destructors associated with `Type` by
removing the internal `Box` and `Vec` indirections. These destructors meant that
a lot of landing pads and extra code were generated to manage the runtime
representations. Instead everything can basically be statically computed and
shoved into rodata, so all we need is a giant string compare to lookup what's
what.

Closes #28273

This code was currently only exercised on OSX, but this applies the same method
of subtraction used on Linux which doesn't have the same overflow issues.

Note that this currently includes no tests, but that's because this is only
visible with debug assertions enabled. Soon, however, I'll enable debug
assertions on all auto builds on the bots so we should get testing for this.

Closes #32268

Refactor rustc_trans::save to allow other backends than CSV

r? @nrc

Things done:
* Moved `(.*)Data` structs to an own module, so they can be imported easily (`use data::*`).
* Created a `Dump` trait with callbacks for dumping items.
* Refactored `DumpCsvVisitor` to use an implementor of `Dump` instead of dumping as CSV. Renamed it to `DumpVisitor`.
* Created a `DumpCsv` struct that implements `Dump` and serializes items as CSV.

I tried to extract some of the logic contained in `FmtStr` and `Recorder`, such as normalization of ids (I put it in `DumpVisitor`). I think it makes sense to provide the same information to other implementors of `Dump`, instead of normalizing only for `DumpCsv`. However, there is still some logic related to spans implemented only for `DumpCsv`. I just thought it would be better to merge this as soon as possible, since there are so much changes, and fix this afterwards.

Move rustc_trans::*Data to own module
Add new data to rustc_trans::save
Create a Dump trait implemented by structs that dump save analysis data
Split dump_csv into a DumpVisitor and a CsvDumper

The LLVM change [1] in #32239 unfortunately broke the LLVM build on MinGW, so
this LLVM submodule update brings in one more fix [2] which should hopefully
remedy that.

Once this lands we should be able to immediately start gating on this to prevent
it from happening again.

[1]: https://github.com/rust-lang/llvm/commit/be89e4b5
[2]: https://github.com/rust-lang/llvm/commit/3dcd2c84

Shorter output for `rustc --test` binaries.

Until now, a program created with `rustc --test` prints at least one line per test. This can be very verbose, especially with [data-driven tests](https://internals.rust-lang.org/t/test-and-external-test-harnesses/3145) when hundreds or thousands of tests is not rare.

This changes the default output to one character per test (except metrics and benchmarks results which have additional data to show):

```
     Running target/debug/wpt-75c594dc1e6e6187

running 314 tests
..............................................................................
..............................................................................
..............................................................................
..............................................................................
..
test result: ok. 314 passed; 0 failed; 0 ignored; 0 measured
```

<s>The previous behavior is available by passing `--verbose` to the test program. Maybe `cargo test --verbose` could be changed to do that?</s> **Edit:** the default is now unchanged, `-q` or `--quiet` enables the new output.