Summary:
The allocator interface is supposed to have 16 byte alignment (to keep
it consistent with the CPU allocator. We could probably drop this to 8
if desires.) But this was not enforced because the number of bytes used
for the bitfield sometimes resulted in alignment of 8 instead of 16.
Explicitly align the number of bytes to be a multiple of 16 even if
unused.

Summary:
This patch uses the actual allocator interface to implement
`aligned_alloc`. We do this by simply rounding up the amount allocated.
Because of how index calculation works, any offset within an allocated
pointer will still map to the same chunk, so we can just adjust
internally and it will free all the same.

Summary:
Now that we have `malloc` we can implement `realloc` efficiently. This
uses the known chunk sizes to avoid unnecessary allocations. We just
return nullptr for NVPTX. I'd remove the list for the entrypoint but
then the libc++ code would stop working. When someone writes the NVPTX
support this will be trivial.

This is the first step in preparation for:
https://discourse.llvm.org/t/rfc-make-clang-builtin-math-functions-constexpr-with-llvm-libc-to-support-c-23-constexpr-math-functions/86450

Summary:
This is the big patch that implements an efficient device-side `malloc`
on the GPU. This is the first pass and many improvements will be made
later.

The scheme revolves around using a global reference counted pointer to
hand out access to a dynamically created and destroyed slab interface.
The slab is simply a large bitfield with one bit for each slab. All
allocations are the same size in a slab, so different sized allocations
are done through different slabs.

Allocation is thus searching for or creating a slab for the desired
slab, reserving space, and then searching for a free bit. Freeing is
clearing the bit and then releasing the space.

This interface allows memory to dynamically grow and shrink. Future
patches will have different modes to allow fast first-time-use as well
as a non-RPC version.

Summary:
The scan operation implemented here only works if there are contiguous
ones in the executation mask that can be used to propagate the result.
There are two solutions to this, one is to enter 'whole-wave-mode' and
forcibly turn them back on, or to do this serially. This implementation
does the latter because it's more portable, but checks to see if the
parallel fast-path is applicable.

Needs to be backported for correct behavior and because it fixes a
failing libc test.

Summary:
These helpers are very useful but currently absent. They allow the user
to get a bitmask representing the matches within the warp. I have made
an executive decision to drop the `predicate` return from `match_all`
because it's easily testable with `match_all() == __activemask()`.

Fixes:


llvm-project/libc/test/integration/src/pthread/pthread_rwlock_test.cpp:59:29:
    warning: missing field '__preference' initializer
    [-Wmissing-field-initializers]
       59 |   pthread_rwlock_t rwlock = PTHREAD_RWLOCK_INITIALIZER;
          |                             ^

Also, add a test that demonstrates the same issue for
PTHREAD_MUTEX_INITIALIZER, and fix that, too.

PTHREAD_ONCE_INIT does not have this issue and does have test coverage.