Fixes rust-lang#36115.

I timed this locally and plain old `gzip` took 2m06s while `gzip -9` took a
whopping 6m23s to save a mere 4MB out of 1.2GB. Let's shave a few minutes off
the Android builder by turning down the compression level.

Turns out that even though all these functions take a `size_t` they don't
actually work that well with anything larger than the maximum value of
`ssize_t`, the return value. Furthermore it looks like OSX rejects any
read/write requests larger than `INT_MAX - 1`. Handle all these cases by just
clamping the maximum size of a read/write on Unix to a platform-specific value.

Closes rust-lang#38590

"verboten" is german for "forbidden"

- `--emit=asm --target=nvptx64-nvidia-cuda` can be used to turn a crate
  into a PTX module (a `.s` file).

- intrinsics like `__syncthreads` and `blockIdx.x` are exposed as
  `"platform-intrinsics"`.

- "cabi" has been implemented for the nvptx and nvptx64 architectures.
  i.e. `extern "C"` works.

- a new ABI, `"ptx-kernel"`. That can be used to generate "global"
  functions. Example: `extern "ptx-kernel" fn kernel() { .. }`. All
  other functions are "device" functions.

We no longer use it, we use sccache

These are no longer used when running containers and tests.

Fixes rust-lang#38485.

On Android we only have one test thread for supposed problems with concurrency
and the remote debugger. Not all of our suites require one concurrency, however,
and suites like compile-fail or pretty can be much faster if they're
parallelized on Travis.

This commit only sets the test threads to one on Android for suites which
actually run code, and other suites aren't tampered with.

Show errors sooner and try not to hide them behind lots of other walls of text.

A new option is introduced under the `[llvm]` section of `config.toml`,
`targets`, for overriding the list of LLVM targets to build support for.
The option is passed through to LLVM configure script. Also notes are
added about the implications of (ab)using the option; since the default
is not changed, and users of the option are expected to know what
they're doing anyway (as every porter should), the impact should be
minimal.

Fixes rust-lang#38200.

(Minor typo fix.)

The "support" in this case is possessed by the "programmer", and that ownership should be indicated by an apostrophe.

Resetting the terminal should first try `sgr0` (as per the comment), not
`sg0` which I believe to be a typo.

This will at least fix rustc output in Emacs terminals (e.g., ansi-term)
with `TERM=eterm-color` which does not provide the next fallback `sgr`. In
such a terminal, the final fallback `op` (`\e[39;49`) is used  which
resets only colors, not all attributes. This causes all text to be
printed in bold from the first string printed in bold by rustc onwards,
including the terminal prompt and the output from all following commands.

The typo seems to have been introduced by rust-lang#29999

(Minor typo fix.)

Since the word `i32` starts with a vowel, the indefinite article should use "an", not "a" \[[1](http://www.dictionary.com/browse/an)\]. (Previously there was one instance of "an i32" and two instances of "a i32", so at least something is wrong!) Since I believe that "an" is the correct form, I aligned everything with that.

This commit switches the rustbuild build system to compiling the
compiler twice for a normal bootstrap rather than the historical three
times.

Rust is a bootstrapped language which means that a previous version of
the compiler is used to build the next version of the compiler. Over
time, however, we change many parts of compiler artifacts such as the
metadata format, symbol names, etc. These changes make artifacts from
one compiler incompatible from another compiler. Consequently if a
compiler wants to be able to use some artifacts then it itself must have
compiled the artifacts.

Historically the rustc build system has achieved this by compiling the
compiler three times:

* An older compiler (stage0) is downloaded to kick off the chain.
* This compiler now compiles a new compiler (stage1)
* The stage1 compiler then compiles another compiler (stage2)
* Finally, the stage2 compiler needs libraries to link against, so it
  compiles all the libraries again.

This entire process amounts in compiling the compiler three times.
Additionally, this process always guarantees that the Rust source tree
can compile itself because the stage2 compiler (created by a freshly
created compiler) would successfully compile itself again. This
property, ensuring Rust can compile itself, is quite important!

In general, though, this third compilation is not required for general
purpose development on the compiler. The third compiler (stage2) can
reuse the libraries that were created during the second compile. In
other words, the second compilation can produce both a compiler and the
libraries that compiler will use. These artifacts *must* be compatible
due to the way plugins work today anyway, and they were created by the
same source code so they *should* be compatible as well.

So given all that, this commit switches the default build process to
only compile the compiler three times, avoiding this third compilation
by copying artifacts from the previous one. Along the way a new entry in
the Travis matrix was also added to ensure that our full bootstrap can
succeed. This entry does not run tests, though, as it should not be
necessary.

To restore the old behavior of a full bootstrap (three compiles) you can
either pass:

    ./configure --enable-full-bootstrap

or if you're using config.toml:

    [build]
    full-bootstrap = true

Overall this will hopefully be an easy 33% win in build times of the
compiler. If we do 33% less work we should be 33% faster! This in turn
should affect cycle times and such on Travis and AppVeyor positively as
well as making it easier to work on the compiler itself.

In an ongoing effort to optimize the runtime of the Android cross builder this
commit updates the pretty test suites to run only for host platforms, not for
target platforms as well. This means we'll still keep running all the suites but
we'll only run them for configured hosts, not for configured targets. This
notably means that we won't be running these suites on Android or musl targets,
for example.

From suggestions at https://users.rust-lang.org/t/what-stable-rust-applications-do-you-use-frequently/7618

This adds some applications which use stable Rust and come with their own
lockfiles in tree.

ripgrep, xsv, and bins have 33 unique dependencies between them.

* Update to ripgrep HEAD because the previous rev would still change the lock
  file when `cargo build` was issued.
* Remove `bins` as it depends on OpenSSL on Windows, which won't work on our
  bots
* Update rev of tokei to get a rev that doesn't change the lockfile

This commit relegates all pretty tests to not get run by default and rather get
run as part of an "aux" test suite. This "aux" suite is renamed from the old
"cargotest" suite to just collect tests that don't need to run everywhere but
should at least pass on Unix/Windows.

Incrementing the `Archive::child_iterator` fetches and validates the next child.
This can trigger an error, which we previously checked on the *next* call to `LLVMRustArchiveIteratorNext()`.
This means we ignore the last error if we stop iterating halfway through.
This is harmless (we don't access the child, after all) but LLVM 4.0 calls `abort()` if *any* error goes unchecked, even a success value.
This means that basically any rustc invocation that opens an archive and searches through it would die.

The solution implemented here is to change the order of operations, such that
advancing the iterator and fetching the newly-validated iterator happens in the same `Next()` call.
This keeps the error handling behavior as before but ensures all `Error`s get checked.

Add some more repos to cargotest

From suggestions at https://users.rust-lang.org/t/what-stable-rust-applications-do-you-use-frequently/7618

This adds some applications which use stable Rust and come with their own lockfiles in their respective trees.

ripgrep, xsv, and bins have 33 unique dependencies between them.

I alphabetized the list by project name because that seems tidier.

r? @brson