[codex] Add sentinel readiness status

[deanbaker]

🤖

Context

We want to support an automated sentinel-drop pipeline that runs once a day during low-traffic windows. That pipeline should only drop _spirit_sentinel when Spirit is actually ready to proceed to cutover.

For multi-day migrations, the sentinel table can exist for a long time while copy/checksum work is still in progress. With the current table-existence-only signal, automation cannot distinguish "sentinel exists because cutover is deferred and ready" from "sentinel exists because the migration is still copying/checksumming." Dropping it too early can remove the operator safety gate before Spirit has reached the cutover wait point.

Summary

• Add a status column to Spirit-created sentinel tables.
• Initialize sentinel rows as status = 'copying'.
• Mark sentinel rows as status = 'ready' only when Spirit reaches the sentinel wait point before cutover.
• Keep legacy/manual sentinel tables compatible: if the sentinel exists without a status column, Spirit still waits for it to be dropped.
• Treat a sentinel dropped during the ready-status update race as a normal release, not a migration failure.
• Document the readiness signal for both migrate and move.

Operator Impact

Automation can now query the sentinel row and only drop the table when it sees ready, making daily low-traffic cutover windows safe for migrations that run across multiple days.

Validation

• go test ./pkg/migration ./pkg/move -run '^$'