fix(tests): health-check wait in serve_integration — unflakes Proptest

[avrabe]

The flake

`api_artifacts_unfiltered` and `api_artifacts_search` in `serve_integration.rs` intermittently fail on `assert_eq!(status, 200)` — parsed status is 0, meaning the HTTP response was empty or unparseable. Happens under Proptest (extended) load (PROPTEST_CASES=1000, high CPU pressure).

Root cause

`start_server()` waits for TCP accept, then returns. But the socket binds before the artifact store finishes loading. Under load, `fetch()` races and hits the server mid-load, getting a closed connection or an error response.

Fix

Poll `/api/v1/health` until it returns HTTP/1.1 200 before declaring the server ready. The health handler only becomes reachable after routing is fully initialized and the store is populated.

Same 30s timeout, same retry cadence — just a stricter readiness check.

Test plan

• `cargo build -p rivet-cli --tests` clean
• CI green — this is the CI fix

This PR blocks nothing it doesn't fix. Other PRs (#157 salsa especially) can rebase to pick this up.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!