OptimizedNodeBridge should not exceed maxProcesses under concurrent load

[bbopen]

Summary

Concurrent calls can race executeRequest and spawn more than maxProcesses workers.

Failure mode

• Multiple requests see processPool.length < maxProcesses and spawn concurrently, overshooting the cap.
• Pool size grows beyond intended limits; resource usage spikes.

Repro (adversarial)

• Configure maxProcesses=2 and fire many concurrent requests at startup.
• Observe pool size exceeding 2.

Acceptance criteria

• Worker spawning is synchronized to respect maxProcesses under concurrency.
• Adversarial test asserts pool size never exceeds cap.

[bbopen]

Fixed in PR #153: WorkerPool uses inFlightCount semaphore with pendingCreations tracking to prevent race conditions. The semaphore is decremented after task completion and worker creation is tracked to avoid exceeding maxProcesses.