Revert async cancellation behaviour to match musl

[sbc100]

This change is partial revert of #19963. We now match musl's behavior
which is that async cancellation does not work to cancel a thread
that is blocked in pthread_mutex_lock.

This reverts a few emscripten-specific patches to some very low level
parts of musl.

[sbc100]

From Gemini:

✦ The current changes revert Emscripten's asynchronous cancellation behavior to match musl's implementation, where pthread_mutex_lock (and timedlock) are not cancellation points, even when
PTHREAD_CANCEL_ASYNCHRONOUS is enabled.

Key Changes Reviewed:

1. system/lib/libc/musl/src/thread/__timedwait.c:

   • Updated comments to better describe the _cp (cancellation point) suffix and how PTHREAD_CANCEL_MASKED behaves (emulating futex syscall returning ECANCELED).
   • Slightly refactored the timeout logic for clarity, returning ETIMEDOUT directly when the sleep duration is exhausted.

2. system/lib/libc/musl/src/thread/pthread_cancel.c:

   • Removed Emscripten-specific #ifdef blocks that previously allowed async cancellation even when canceldisable was set. This aligns the logic with musl's standard behavior where cancellation is
     strictly governed by canceldisable.

3. system/lib/pthread/emscripten_futex_wait.c:

   • Refined the cancelable flag logic: bool cancelable = !self->canceldisable && self->cancelasync;. This ensures that low-level futex waits only honor async cancellation when it's explicitly enabled
     and not masked.

4. test/pthread/test_pthread_cancel_async.c:

   • Updated the test to verify that pthread_mutex_timedlock now correctly times out (returning ETIMEDOUT) instead of being interrupted by the cancellation request.
   • Added a pthread_testcancel() call after the timeout to confirm the thread is eventually cancelled at a valid cancellation point.

5. test/test_posixtest.py:

   • Added test_pthread_setcanceltype_1_1 to the unsupported_noreturn list, noting that async cancellation within pthread_mutex_lock is no longer supported (matching musl).

6. Code Size Updates:

   • Minor increases (~5 bytes in WASM) in test_codesize_minimal_pthreads.json and its memory-growth variant, likely due to the slightly more robust check in emscripten_futex_wait.

Conclusion:
The implementation is architecturally sound and achieves the goal of musl-compatibility for thread cancellation. The tests have been updated to reflect the new behavior, and the code size impact is
negligible.

[sbc100]

This change paves the way for the final simplification in this series #26550