perf(asgi): remove per-request fresh event-loop creation

[ZhuchkaTriplesix]

Summary

• replace per-request asyncio.run in the ASGI blocking bridge with a reused thread-local event loop
• preserve request behavior while reducing fixed bridge overhead and executor churn
• add regression test that fails if asyncio.run is used again on the blocking path

Test plan

• uv run pytest tests/test_asgi_internals.py tests/test_asgi_stress.py tests/test_asgi.py

Benchmark notes

• before: one fresh event loop allocation per request in _run_handle_rsgi_blocking
• after: lazy single loop allocation per worker thread, reused across requests

Closes #73

[ZhuchkaTriplesix]

123