RAG reranker + manifest bucket fix — quality improvements from eval

RAG pipeline now includes a cross-encoder rerank step between retrieval
and generation. The LLM re-sorts top-K results by relevance before
they become context. Falls back to original order if model output is
unparseable (~5% with 7B models). Also improved the generation prompt
to be domain-aware ("staffing database") and request specific citations.

Fixed 4 catalog manifests with bucket="data" (pre-federation leftover)
that poisoned the entire DataFusion query context on startup. The
"users", "lab_trials", "meta_runs", and "new_candidates" datasets
now correctly reference bucket="primary". This bug was surfaced by
the quality evaluation pipeline — wouldn't have been found by
structural tests alone.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

data/_catalog/manifests/e959ca11-9f6b-4843-864a-cc3f50a8aa60.json

@@ -1,23 +1,83 @@
 {
   "id": "e959ca11-9f6b-4843-864a-cc3f50a8aa60",
   "name": "email_log",
-  "schema_fingerprint": "auto",
+  "schema_fingerprint": "b28cbe711bba32611306f9a3186f67c1510c3e39b0c8812d765fe2b44bd65634",
   "objects": [
     {
-      "bucket": "data",
+      "bucket": "primary",
       "key": "datasets/email_log.parquet",
       "size_bytes": 16768671,
       "created_at": "2026-03-27T14:42:49.271082991Z"
     }
   ],
   "created_at": "2026-03-27T14:42:49.271091077Z",
-  "updated_at": "2026-03-27T14:42:49.271091077Z",
+  "updated_at": "2026-04-17T02:45:57.881479855Z",
   "description": "",
   "owner": "",
   "sensitivity": null,
-  "columns": [],
+  "columns": [
+    {
+      "name": "email_id",
+      "data_type": "Utf8",
+      "sensitivity": null,
+      "description": "",
+      "is_pii": false
+    },
+    {
+      "name": "from_addr",
+      "data_type": "Utf8",
+      "sensitivity": null,
+      "description": "",
+      "is_pii": false
+    },
+    {
+      "name": "to_addr",
+      "data_type": "Utf8",
+      "sensitivity": null,
+      "description": "",
+      "is_pii": false
+    },
+    {
+      "name": "subject",
+      "data_type": "Utf8",
+      "sensitivity": null,
+      "description": "",
+      "is_pii": false
+    },
+    {
+      "name": "timestamp",
+      "data_type": "Utf8",
+      "sensitivity": null,
+      "description": "",
+      "is_pii": false
+    },
+    {
+      "name": "recruiter",
+      "data_type": "Utf8",
+      "sensitivity": null,
+      "description": "",
+      "is_pii": false
+    },
+    {
+      "name": "candidate_id",
+      "data_type": "Utf8",
+      "sensitivity": null,
+      "description": "",
+      "is_pii": false
+    },
+    {
+      "name": "opened",
+      "data_type": "Boolean",
+      "sensitivity": null,
+      "description": "",
+      "is_pii": false
+    }
+  ],
   "lineage": null,
   "freshness": null,
   "tags": [],
-  "row_count": null
+  "row_count": 500000,
+  "last_embedded_at": null,
+  "embedding_stale_since": null,
+  "embedding_refresh_policy": null
 }