🐛 Make GlobalThreadPool initialization thread-safe

[original4422]

Summary

• synchronize lazy singleton construction and publish the pool only after its executor is ready
• retain the module-level five-worker pool and warn when a later caller requests a different worker count
• add deterministic CPU-only regression coverage for construction races, conflicting worker counts, and submit delegation

Root cause

GlobalThreadPool.__new__ published _instance before ThreadPoolExecutor construction completed, so another thread could receive an incompletely initialized singleton. It also silently retained the first worker count on later conflicting calls.

Testing

• uv run --with pytest pytest test/backend/utils/test_thread_utils.py -q (3 passed)
• uv run --with ruff ruff check backend/utils/thread_utils.py test/backend/utils/test_thread_utils.py
• git diff --check

Fixes #3820

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.