chore: update verisimdb training data from Hypatia test runs

Updated recipe confidences, outcomes, and health data from neural training
pipeline execution. index.json total_scans synced to 301.

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

Author: hyperpolymath

verisimdb/verisimdb-data/health/hypatia.json

@@ -1,5 +1,5 @@
 {
-  "checked_at": "2026-03-16T00:52:39.833241Z",
+  "checked_at": "2026-03-16T00:57:42.896090Z",
   "checks": {
     "circuit_breaker": {
       "failures": 0,
@@ -10,14 +10,14 @@
       "status": "pass"
     },
     "fleet_access": {
-      "last_modified": "{{2026, 3, 16}, {0, 52, 38}}",
-      "size_bytes": 1602032,
+      "last_modified": "{{2026, 3, 16}, {0, 57, 42}}",
+      "size_bytes": 1685837,
       "status": "pass"
     },
     "learning_scheduler": {
       "confidence_updates": 6,
-      "last_run": "~U[2026-03-16 00:52:41.065045Z]",
-      "outcomes_processed": 21253,
+      "last_run": "~U[2026-03-16 00:57:46.852156Z]",
+      "outcomes_processed": 21817,
       "status": "pass"
     },
     "recipe_health": {