Since LLVM doesn't vectorize the loop for us, do unaligned reads
of a larger type and use LLVM's bswap intrinsic to do the
reversing of the actual bytes.  cfg!-restricted to x86 and
x86_64, as I assume it wouldn't help on things like ARMv5.

Also makes [u16]::reverse() a more modest 1.5x faster by
loading/storing u32 and swapping the u16s with ROT16.

Thank you ptr::*_unaligned for making this easy :)

None of these are affected by e8fad32.

Part of rust-lang#29378

Part of rust-lang#29378

- Adds an explanantion of what `park` does in the `unpark` documentation.
- Adds a link to the module doc.

Part of rust-lang#29378

Part of rust-lang#29378

- Moves the module documentation into `park`.
- Add the same example as the one from `unpark` to `park`.

The best way to do this wasn't in the documentation, and the ways that
were there needed some extra text to elaborate.

Fixes rust-lang#40159

This is something I always forget, so let's actually
explain in the docs.

When `RUST_BACKTRACE=1`, remove all frames after
`__rust_maybe_catch_panic`. Tested on `main`, threads, tests and
benches. Cleaning of the top of the stacktrace is let to a future PR.

Fixes rust-lang#40201

See rust-lang#41815

Add more ways to create a PathBuf to docs

The best way to do this wasn't in the documentation, and the ways that
were there needed some extra text to elaborate.

Fixes rust-lang#40159

/cc @nagisa

…ushi

Improve docs on Arc<T> and Send/Sync

This is something I always forget, so let's actually
explain in the docs.

I didn't fully link up everything here, but I'd like to make sure that the wording is okay before I bother.

Make [u8]::reverse() 5x faster

Since LLVM doesn't vectorize the loop for us, do unaligned reads of a larger type and use LLVM's bswap intrinsic to do the reversing of the actual bytes.  cfg!-restricted to x86 and x86_64, as I assume it wouldn't help on things like ARMv5.

Also makes [u16]::reverse() a more modest 1.5x faster by loading/storing u32 and swapping the u16s with ROT16.

Thank you ptr::*_unaligned for making this easy :)

Benchmark results (from my i5-2500K):
```text
test slice::reverse_u8      ... bench:  273,836 ns/iter (+/- 15,592) =  3829 MB/s
test slice::reverse_u16     ... bench:  139,793 ns/iter (+/- 17,748) =  7500 MB/s
test slice::reverse_u32     ... bench:   74,997 ns/iter  (+/- 5,130) = 13981 MB/s
test slice::reverse_u64     ... bench:   47,452 ns/iter  (+/- 2,213) = 22097 MB/s

test slice::reverse_u8      ... bench:   52,170 ns/iter (+/- 3,962) = 20099 MB/s
test slice::reverse_u16     ... bench:   93,330 ns/iter (+/- 4,412) = 11235 MB/s
test slice::reverse_u32     ... bench:   74,731 ns/iter (+/- 1,425) = 14031 MB/s
test slice::reverse_u64     ... bench:   47,556 ns/iter (+/- 3,025) = 22049 MB/s
```

If you're curious about the assembly, instead of doing this
```
movzx	eax, byte ptr [rdi]
movzx	ecx, byte ptr [rsi]
mov	byte ptr [rdi], cl
mov	byte ptr [rsi], al
```
it does this
```
mov	rax, qword ptr [rdx]
mov	rbx, qword ptr [r11 + rcx - 8]
bswap	rbx
mov	qword ptr [rdx], rbx
bswap	rax
mov	qword ptr [r11 + rcx - 8], rax
```

[DOC] Improve the thread::park and thread::unpark documentation

Part of rust-lang#29378 .

Takes care of the documentation for `park`, `park_duration` and also improves the `unpark` example.

- `park should` have its module documentation inlined here, and cleaned up.
- `park_timeout` could use links to `park`.

…r=alexcrichton

Improve cleaning of the bottom of the backtrace

Following rust-lang#40264. It only cleans the bottom of the trace (after the main). It handles correctly the normal main, tests, benchmarks and threads.

I kept `skipped_before` since it will be used later for the cleaning of the top.

…ichton

Update jemalloc to 4.5.0

* pointing jemalloc submodule to the 4.5.0 HEAD + generated configure script
* removed the now unnecessary extra config for windows. jemalloc is now much better at configuring itself overall.

Let's see what the CI think.

Improve E0477 error message

Part of rust-lang#41816.

r? @nagisa

…alexcrichton

ci: Update android ndk and sdk

Make install-sdk.sh and install-ndk.sh more generic so future updates can be made directly on Dockerfile. Update ndk to r13b, which will be necessary to make host builds for android (in the future). Update sdk to r25.2.5 (maybe some emulator performance improvement).

…hton

Add disabled android host builders

Introduce the concept of disabled builder. A disabled builder is one that is not run by travis. It is intended to be run by the user who wants a rustc for a tier 2 or 3 platform. Off corse, there is no guarantee that it will work.