fix(tests): object-store retention-prune tests no longer expire with the wall clock (#1066)

[masukai]

Summary

Closes #1066. Fixes CI, currently broken on main for every PR (confirmed independently on a clean checkout).

test_remote_history_applies_entry_cap_only_during_prune and test_prune_caps_by_started_at_not_append_order hardcoded August 2026 dates and asserted an exact prune(retention_days=30) removed-count against ObjectStoreHistoryStore._pruned()'s real datetime.now() cutoff. As calendar time advanced past the 30-day window, a second fixture entry started aging out unintentionally — these tests were meant to isolate max_entries count-capping from age-based retention (per the first test's own name), and the real-clock dependency broke that isolation.

Fix

Dates are now computed relative to the actual test-run time (now - timedelta(days=N), all within a few days — never near the 30-day boundary) instead of hardcoded absolutes. Audited the rest of the file plus local history/DLQ tests for the same pattern — no other instances found.

Verification

• ruff check drt tests / mypy drt — clean
• pytest tests/unit/test_state_objectstore.py — 29 passed
• pytest tests/unit — full suite green except 2 known-unrelated test_cli_version.py failures (long-scratchpad-path artifact)

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!