Remove timeout slicing from __timedwait. NFC

[sbc100]

This removes the timeout slicing from __timedwait and instead relies
on the existing slicing in emscripten_futex_wait.c.

This is a followup to #26511 and #26471 (which did the same for the
slicing in __wait).

[sbc100]

This change is currently stacked on top of #26549

[sbc100]

From Gemini:

✦ Review of Commit 0050751: "Remove timeout slicing from __timedwait. NFC"

This commit simplifies the Emscripten-specific wait logic by centralizing timeout slicing in emscripten_futex_wait.c.

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

• Change: Removed the manual slicing loop (max_ms_slice_to_sleep) and the cancellation check within __timedwait_cp. It now directly calls emscripten_futex_wait.
• Assessment: High-quality simplification. By delegating the slicing and cancellation handling to the lower-level emscripten_futex_wait, we reduce code duplication and complexity in the musl-specific
  part of the library.
• Naming: Changed msecsToSleep (camelCase) to msecs_to_sleep (snake_case), which is more consistent with musl/libc style.

2. system/lib/pthread/emscripten_futex_wait.c

• Change: Updated cancelable calculation to self->canceldisable != PTHREAD_CANCEL_DISABLE.
• Assessment: This is a critical change. It ensures that any thread that hasn't explicitly disabled cancellation is considered "cancelable" for the purpose of the slicing loop. This correctly includes
  PTHREAD_CANCEL_MASKED (used by pthread_cond_wait), which is essential for correct condition variable behavior.
• Refinement: Replaced a simple break with a more robust error return:

1 emscripten_conditional_set_current_thread_status(EM_THREAD_STATUS_WAITFUTEX, EM_THREAD_STATUS_RUNNING);
2 return -ECANCELED;
This ensures the thread status is correctly restored before returning the cancellation error to the caller (e.g., __timedwait_cp).

3. Code Size (test/codesize/*.json)

• Change: Significant reduction in WASM size (~200 bytes).
• Assessment: Excellent. Centralizing logic often leads to smaller binaries in Emscripten, and this is a clear win. The reordering of imports (exit vs emscripten_get_now) is a side effect of the symbol
  usage changes but is harmless.

4. Structural Consistency

• This change follows the pattern established in Remove some of the complexity from __timedwait.c. NFC #26511 and Move sleep slicing from musl's __wait to the lower level emscripten_futex_wait. NFC #26471 for __wait.c. Consistently applying this architecture across all wait primitives makes the threading library much easier to maintain.

Final Verdict: LGTM
The change is a clean, "No Functional Change" (NFC) refactor that improves maintainability and reduces code size without altering the intended behavior (now that async cancellation has been correctly
reverted in the previous commit).