test: deflake integration tests by polling instead of fixed sleeps

[vdusek]

Summary

Started as a fix for this flaky CI run of test_request_queue_unlock_requests[sync] (assert 2 == 3 on unlocked_count, caused by replication lag between list_and_lock_head and unlock_requests), and grew into deflaking the integration test suite as a whole.

Changes:

• Add a poll_until_condition helper to tests/integration/_utils.py: polls a sync-or-async callable at a constant interval until a condition holds or a wall-clock timeout expires. An optional backoff_factor multiplies the interval after each poll, for highly variable waits (e.g. Actor run container startup) where a growing delay covers a long timeout with few calls.
• Fix the flaky unlock test by polling list_head until the locked IDs disappear from the queue head before unlocking.
• Replace all hand-rolled for _ in range(5): sleep(1); read; break polling loops (10×) and single fixed sleep(1) waits (26×) across the request queue, dataset, key-value store, and run integration tests with poll_until_condition. This makes the tests both faster on the happy path (no unconditional sleep) and more robust under load (polls until the timeout instead of hoping 1 s is enough).
• Generalize maybe_await to accept any awaitable.

The three iterate_keys sleeps in the KVS tests are intentionally left as-is: draining an iterator per attempt wants attempt-count semantics (like collect_iterate_until_present), not a wall-clock deadline.

Follow-up to #786.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.82%. Comparing base (d2ab2ae) to head (c856ab1).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #844      +/-   ##
==========================================
+ Coverage   94.68%   94.82%   +0.13%     
==========================================
  Files          48       48              
  Lines        5045     5045              
==========================================
+ Hits         4777     4784       +7     
+ Misses        268      261       -7     

Flag	Coverage Δ
integration	93.24% <ø> (+0.23%)	⬆️
unit	83.52% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.