Size the recovery-test time budgets for slow CI runners (flake fix, tests only)

[seathatflowsinourveins]

Summary

• Fixes intermittent hosted-CI failures in tests/test_adaptive_paper_recovery.py. Two tests hit 'needs_attention' != 'passed': test_admission_freeze_does_not_block_confirmed_owned_exit (seen on Require the promotion-gate result to report every named gate check (SYN) #77) and test_partial_buy_cancel_then_exact_nine_decimal_exit_after_expired_trial (seen on Let catalog-freshness open a reviewable, off-by-default evidence PR #95, reported by a peer session).
• Cause: the test's fake port publishes a single SPY quote at start(), and the fixture set quote_max_age_seconds = 0.05. Recovery requires a fresh quote before the owned exit, and the snapshot and ledger writes in between can take longer than 50 ms on a slow CI disk. The quote then goes stale, recovery_quote_not_fresh is raised, and the result is needs_attention.
• Reproduced deterministically by injecting an 80 ms delay into the fake snapshot. With the old budgets both tests fail (2/2); with the new budgets both pass (0/2).
• Fix (test fixture only; engine code unchanged): the fixture now allows a 1 s quote age and a 3 s recovery deadline. The deadline test sets its own 1 s budget. The 30-second-stale quote case still fails as intended.

Evidence

• tests.test_adaptive_paper_recovery: 21 OK under both interpreters.
• python3 -m unittest: 2624 OK (330 skipped).
• validate.py, validate_catalogs.py, trading_gates.py --check, evidence_manifest.py --check and build_ecosystem.py --check all rc=0.
• Guarded gitleaks: no leaks.

🤖 Generated with Claude Code