feat: auto-build training dataset from DB annotations before training

- Add build_dataset_from_db() that exports candles from DB, runs feature
  engineering, and ingests span annotations into labeled CSV
- Call it automatically in _run_training_background before training starts
- Add POST /training/build-dataset endpoint for standalone use
- Add Next.js proxy route /api/training/build-dataset
- Update TrainingPanel: remove dataset-missing block on Start Training,
  show informational message that dataset builds automatically

src/components/TrainingPanel.tsx

@@ -228,8 +228,7 @@ export default function TrainingPanel() {
     }
   };
 
-  const datasetMissing = datasetInfo !== null && !datasetInfo.exists;
-  const canTrain = !isTraining && !datasetMissing && datasetInfo !== null;
+  const canTrain = !isTraining;
 
   return (
     <div>
@@ -272,8 +271,8 @@ export default function TrainingPanel() {
                 )}
               </>
             ) : (
-              <p className="text-orange-500">
-                No training dataset found. Export annotations first.
+              <p className="text-muted-foreground">
+                No cached dataset. It will be built automatically from your annotations when training starts.
               </p>
             )}
           </div>