test(memory): stop daemon before direct database reads

[ScriptedAlchemy]

Summary

• stop and fully drop the fixture daemon before the memory eval harness opens SQLite directly
• on Windows, wait until the recorded loopback endpoint is no longer connectable before restarting
• restart the daemon only after the direct connection has been dropped
• apply the lifecycle boundary to both scalar assertions and seeded-source snapshots

Root cause

Windows shard 1 in release PR #497 failed because eval_memory_no_pollution queried the fixture database while the daemon still owned a write transaction. The first fix made shard 1 pass, then exposed a Windows teardown race in shard 4: a listener inherited by a child could remain briefly connectable after the owned daemon process exited. The harness now observes endpoint closure rather than assuming process exit immediately releases every inherited handle.

Evidence

• original database-lock failure: https://github.com/ScriptedAlchemy/tracedecay/actions/runs/30903862937/job/91975643965
• teardown race exposed by first fix: https://github.com/ScriptedAlchemy/tracedecay/actions/runs/30906452828/job/91983721219
• cargo test --test memory_suite memory_eval_test::eval_memory_no_pollution -- --exact --nocapture
• cargo test --test memory_suite (52 passed)
• cargo fmt --all -- --check
• git diff --check

[changeset-bot]

⚠️ No Changeset found

Latest commit: e7d7699

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.