Fix async export deletion test flake

[andystaples]

Summary

• wait for the asynchronously signaled export entity deletion before asserting the job is gone
• use bounded polling instead of an immediate read or fixed sleep

Validation

• python -m pytest tests/durabletask/extensions/history_export/test_client.py -q
• repeated test_async_client_create_list_wait_and_delete 10 times
• python -m flake8 tests/durabletask/extensions/history_export/test_client.py

[Copilot]

Pull request overview

This PR addresses a flaky async history export deletion test by replacing an immediate post-delete read with bounded polling, ensuring the export entity deletion signal has been processed before asserting the job is gone.

Changes:

• Add asyncio import for async-friendly polling and timing.
• After await job_client.delete(), poll await job_client.describe() until it returns None or a deadline is reached (fail-fast with a clear timeout error).

[berndverst]

Reviewed the async deletion race at 075808d. SignalEntity acknowledges enqueueing rather than completed entity processing, so the prior immediate read was racy. The replacement polls the same observable deletion condition with a monotonic 5-second bound and 50 ms async sleeps, and still fails if deletion never occurs. No production/API compatibility change. Focused test passed repeatedly, including a delayed-deletion probe, and the changed test passes flake8.