feat: Add worker health check and auto-restart logic

[vominh1919]

Fixes #41

Problem

If the validation engine worker thread crashes (e.g., due to a key error or OOM), there was no recovery mechanism. The engine would fail permanently.

Solution

• Add _is_worker_alive() method to check if worker process is still running
• Add _restart_worker() method to cleanly restart crashed workers:
  • Terminates old worker process
  • Clears input queue of stale items
  • Fails pending validations with CRASHED status
  • Starts new worker process
• Add _restart_lock to prevent concurrent restart attempts
• Add _pending_validations dict to track in-flight validations
• Update validate_entries() and validate_grouped_entries() to:
  • Check worker health before starting validation
  • Monitor worker health while waiting (5-second intervals)
  • Handle worker crashes gracefully with CRASHED status

Testing

The changes are backwards compatible and maintain the existing API.