feat(services): add async locks to prevent race conditions in graph loading

[jayadityashah]

Description

Resolves connection errors under high concurrent load and enables safe testing with 100+ simultaneous threads.

Type of Change

• feat: New feature
• fix: Bug fix
• docs: Documentation changes
• style: Code style/formatting
• refactor: Code refactoring
• perf: Performance improvement
• test: Tests added/updated
• chore: Maintenance (dependencies, build, etc.)
• ci: CI/CD changes

Related Issues

Not sure if there is a related issue

Changes Made

• Add instance-level locks to prevent concurrent graph loading issues
• Implement thread-safe lock creation with master lock pattern
• Add double-check pattern after acquiring locks for cache consistency
• Include proper cleanup in invalidate_cache() to prevent memory leaks
• Use lazy initialization to avoid event loop issues during startup

Testing

• Unit tests added/updated
• Integration tests added/updated
• E2E tests added/updated
• Manual testing performed

Checklist

• Code follows project style (Ruff formatting)
• All linting checks pass (make lint)
• Type checking passes (make type-check)
• All tests pass (make test)
• Documentation updated (if needed)
• Commit messages follow Conventional Commits
• PR title follows Conventional Commits format

Screenshots (if applicable)

Additional Notes

[codecov-commenter]

Codecov Report

❌ Patch coverage is 86.66667% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/agent_server/services/langgraph_service.py	86.66%	4 Missing ⚠️

📢 Thoughts on this report? Let us know!

[jayadityashah]

@ibbybuilds the 4 lines above that codecov mentioned are my changes, which is why they are different I ran the tests and they worked.

Happy to change discuss more if this commit needs more work.

[jayadityashah]

Does this mean that we write a test for this in the startup like the other tests? And then if the test covers this line, codecov will accept it?

[ibbybuilds]

@jayadityashah coverage above 85% is totally fine, we don't always need 100% coverage. I will review the code and ask for changes if needed, just a little busy these days