P3: Complete Phase 3 Task Registry + Persistence
Implements the 14-table task-store schema from plan §4 with both SQLite
and Redis backends, enabling pod restart resilience and multi-replica HA.
## Changes
- SqliteTaskStore: Full TaskStore trait implementation for all 14 tables
- Tables 1-7: tasks, node_settings_version, aliases, sessions,
idempotency_cache, jobs, leader_lease
- Tables 8-14: canaries, canary_runs, cdc_cursors, tenant_map,
rollover_policies, search_ui_config, admin_sessions
- WAL mode + busy_timeout for concurrent access
- Idempotent migrations with schema version tracking
- RedisTaskStore: Complete TaskStore trait implementation
- Mirrors SQLite keyspace with hash + _index pattern for O(1) lookups
- Uses SET NX/EX for leader leases, ZADD for canary runs
- Pub/Sub for instant admin session revocation
- Rate limiting helpers (search_ui, admin_login with backoff)
- CDC overflow buffer with byte tracking
- Schema migrations: 3-migration system (001_initial, 002_feature_tables,
003_task_registry_fields)
- Tests:
- SQLite: 36 tests including property tests (proptest)
- Redis: 20+ integration tests using testcontainers
- Restart resilience: tasks survive DB close/reopen cycles
- Helm validation: values.schema.json enforces replicas > 1 requires
taskStore.backend: redis
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
63a9207051