fix(db): pause call-log rotate on SQLITE_CORRUPT

[RaviTharuma]

Summary

• On SQLITE_CORRUPT / malformed pager / SQLITE_IOERR during call-log rotation, pause further rotate writes instead of retrying DELETEs against a broken file.
• Record the failure so /api/db/health reports integrity_check_failed (Pager reported SQLITE_CORRUPT during call-log-rotate: …).
• Ordinary filesystem errors still log and do not pause rotation.

Does not auto-REINDEX. That remains unsafe on a live single-writer file.

Related Issues

• Closes the fail-closed gap in fix(db): SQLite integrity_check fails (rowid out of order); call-log rotate SQLITE_CORRUPT #10736 (owner already scoped this as the narrow safe fix)
• Related to docs(docker): document SQLite single-replica HA limits #10817 (single-replica SQLite HA limits)

Validation

• Change type: DB
• Focused tests: node --import tsx/esm --test tests/unit/call-log-rotate-corrupt.test.ts (4/4)
• tests/unit/call-log-file-rotation.test.ts swallow-filesystem case still passes
• npm run lint (CI)
• Branched from origin/release/v3.8.50 (266e39d36)
• New automated test in this PR

Tests Added Or Updated

• tests/unit/call-log-rotate-corrupt.test.ts

Coverage Notes

• src/lib/db/healthCheck.ts pager-note helpers + src/lib/usage/callLogRotation.ts fence covered by the new file.
• One existing orphan-cleanup assertion in call-log-file-rotation.test.ts failed on this machine because better-sqlite3 is built for NODE_MODULE_VERSION 137 vs local 147 (sql.js fallback). Unrelated to this fence; CI image has a matching native module.

Reviewer Notes

• Public /healthz stays process-lifecycle only (by design). Degraded DB is /api/db/health.
• Production evidence for fix(db): SQLite integrity_check fails (rowid out of order); call-log rotate SQLITE_CORRUPT #10736 (real integrity_check, Longhorn block PVC) is on the issue.