Fix Container Shutdown Race Condition

[chris-sanders]

Summary

• Fixed Python runtime error during container shutdown by removing sys.exit(0) from signal handler
• Added global shutdown flag coordination to allow graceful exit without forcing immediate termination
• Created comprehensive test suite to verify fix and prevent regression

Problem

Container shutdown was failing with:

Fatal Python error: _enter_buffered_busy: could not acquire lock for <_io.BufferedReader name='<stdin>'> 
at interpreter shutdown, possibly due to daemon threads

Solution

The race condition was caused by calling sys.exit(0) in the signal handler while logging operations were still in progress. The fix:

1. Removed sys.exit(0) from the signal handler
2. Added a global _shutdown_requested flag that the main loop can check
3. Allow Python to exit naturally after cleanup completes

Test Plan

• Created reproduction test that verifies the race condition is fixed
• Added comprehensive signal handling integration tests
• Added e2e container shutdown reliability tests
• All tests pass locally
• Code quality checks pass (black, ruff, mypy)