Make route deployment and migration retries recover safely

[squarepots]

Summary

• audit remote Route state before every deployment attempt
• persist state=deploying before remote mutation so interrupted attempts have a durable retry checkpoint
• adopt verified in-sync remote state locally without repeating deployment
• allow remote mutation only when audit proves the managed service is missing and local state is not already deployed
• report remote deployment, local state persistence, and observed-state persistence as distinct partial states
• let migration retries reuse the same audit-first deployment contract

Tests

• deployment mutation is allowed only for known safe remote states
• the deploying retry checkpoint survives a fresh state load
• verified in-sync recovery adopts local state without another remote mutation
• migration resumes after an undetermined deployment result without duplicating replacement state

Version impact

patch — fixes existing deployment and migration retry behavior.

Closes #7