Ensure training memory monitor cleans up callbacks

[lukifer23]

Summary

• extend the memory monitor helpers to report start/stop success and expose a callback removal API
• teach train_comprehensive to track monitor lifecycle, unregister its alert callback, and stop monitoring during cleanup
• add a regression test that runs training twice to verify callbacks are not duplicated and the monitor thread stops

Testing

• pytest tests/test_train_memory_monitor.py

---

https://chatgpt.com/codex/tasks/task_e_68d1c2a371b88323861a41c2fc7d9bc8

[Copilot]

Pull Request Overview

This PR implements proper cleanup for the memory monitoring system to prevent resource leaks and callback accumulation. The training function now tracks the memory monitor's lifecycle and ensures proper shutdown when training completes or fails.

• Extended memory monitor API to report start/stop success and expose callback removal functionality
• Modified train_comprehensive to properly manage monitor lifecycle and clean up registered callbacks
• Added comprehensive regression test to verify no callback duplication and proper thread cleanup

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
tests/test_train_memory_monitor.py	New test file with regression test for memory monitor cleanup
azchess/utils/memory_monitor.py	Added return values to start/stop methods and callback removal API
azchess/utils/init.py	Exported new remove_memory_alert_callback function
azchess/training/train.py	Added monitor lifecycle tracking and cleanup in train_comprehensive

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

This generator will never yield any values since the while condition is False. This should be while True: or use a different approach to create an empty generator.

Suggested change

	while False:
	yield None
	if False:
	yield

[Copilot]

[nitpick] This condition check is confusing. Consider simplifying to check if there's nothing to stop: if not self.is_monitoring and (not self.monitor_thread or not self.monitor_thread.is_alive()):

Suggested change

	if not self.is_monitoring and not (self.monitor_thread and self.monitor_thread.is_alive()):
	if not self.is_monitoring and (not self.monitor_thread or not self.monitor_thread.is_alive()):