Fix INVALID_PARAMETER crash in ThreadPoolSchedulerWin::Post by adding null pointer guards

[Copilot]

This PR fixes a Watson failure crash caused by INVALID_PARAMETER (c000000d) in Microsoft.ReactNative.dll!Mso::ThreadPoolSchedulerWin::Post. The crash occurs when ::SubmitThreadpoolWork() is called with a null or invalid TP_WORK handle.

Root Cause

The crash happens when:

1. CreateThreadpoolWork() fails during construction and returns nullptr
2. The object is used during/after destruction
3. Memory corruption affects the m_threadPoolWork member

The stack trace shows the crash originates from line 157 in threadPoolScheduler_win.cpp where ::SubmitThreadpoolWork(m_threadPoolWork.get()) is called without validating the handle.

Changes Made

Added defensive null pointer checks in four critical locations:

1. Post() method: Guards SubmitThreadpoolWork() call and reverts thread count increment if work cannot be submitted
2. AwaitTermination() method: Guards WaitForThreadpoolWorkCallbacks() call to prevent similar crashes
3. WaitForThreadPoolWorkCompletion() method: Guards against null handles in tracked work loop
4. TrackThreadPoolWork() method: Prevents tracking of invalid handles

Example of the fix

Before:

void ThreadPoolSchedulerWin::Post() noexcept {
  // ... thread count logic ...
  ::SubmitThreadpoolWork(m_threadPoolWork.get()); // Can crash if null
}

After:

void ThreadPoolSchedulerWin::Post() noexcept {
  // ... thread count logic ...
  if (m_threadPoolWork.get() != nullptr) {
    ::SubmitThreadpoolWork(m_threadPoolWork.get());
  } else {
    --m_usedThreads; // Revert thread count if we can't submit work
  }
}

These changes ensure Windows ThreadPool APIs are never called with invalid parameters while maintaining proper state consistency.

Fixes #15099.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

Microsoft Reviewers: Open in CodeFlow