Graceful fallback from update-task-v2 to v1 when server doesn't support it

[v1r3n]

Summary

• Auto-detect when the server doesn't support the POST /tasks/update-v2 endpoint (HTTP 404 or 501) and seamlessly fall back to the standard POST /tasks endpoint for the lifetime of the runner
• On fallback, the tight execute loop exits and the worker resumes normal poll-based operation — no task results are lost
• Detection happens once per runner on the first failed v2 call; all subsequent updates go directly to v1 with no overhead
• Added unit tests (11) covering both sync TaskRunner and async AsyncTaskRunner: fallback on 404/501, subsequent calls skip v2, v2 success path unchanged, non-404 errors don't trigger fallback, v1 retry on
  fallback failure
• Added integration test that runs real workflows against the server and verifies completion regardless of v2 availability

Test plan

• Unit tests: python -m pytest tests/unit/automator/test_v2_fallback.py -v (11 passed)
• Existing tests: python -m pytest tests/unit/automator/ -v (235 passed, no regressions)
• Integration perf test: python -m pytest tests/integration/test_update_task_v2_perf.py -v -s (1000 workflows, p50 queue wait 26ms)
• Integration fallback test: python -m pytest tests/integration/test_v2_fallback_intg.py -v -s (against server without v2 support)