Implement reconciliation job for dual-write dirty IDs

[tphakala]

Background

During dual-write mode, if V2 delete fails after legacy deletion succeeds, the detection ID is marked as "dirty" for later reconciliation.

Current State

• Location: internal/datastore/v2/repository/dual_write.go:283
• The dirty ID tracking works for migration phase failures
• No periodic reconciliation job exists for dual-write failures during live operation

Proposed Solution

1. Implement a periodic "Retry Dirty IDs" task that:

   • Scans for dirty IDs
   • Attempts to reconcile V2 state with legacy
   • Clears dirty flag on success
   • Logs failures for manual investigation

2. Consider running this:

   • On a configurable interval (e.g., hourly)
   • At startup to catch any pending dirty IDs

Acceptance Criteria

• Periodic reconciliation job implemented
• Configurable interval
• Proper logging and error handling
• Tests for reconciliation logic