P2.5: Implement task ID reconciliation and /tasks endpoints
Implements plan §3 "Task ID reconciliation":
- Every write fan-out collects per-node taskUid values
- Generate Miroir task ID mtask-<uuid>
- Persist mtask → {node_id: node_task_uid} in in-memory task registry
- Return mtask-xxxxx to client as {"taskUid": ...} in Meilisearch shape
- GET /tasks/{mtask_id} polls every mapped node task, aggregates status
- succeeded: all nodes report succeeded
- failed: any node reports failed; includes per-node error detail
- processing: otherwise
- GET /tasks with Meilisearch-compatible filters (statuses, types, indexUids, from, limit)
- DELETE /tasks/{mtask_id} for best-effort cancellation
Details:
- Polling cadence: exponential backoff (25ms → 50 → 100 → ... → 1s cap)
- In-memory registry using Arc<RwLock<HashMap<String, MiroirTask>>>
- NodeClient trait extended with get_task_status method
- TaskStatusResponse with to_node_status() conversion
- Background polling spawned per task with tokio::spawn
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
aa1982006e