fix(ml): complete ML pipeline fixes and setup

- Fix CCI indicator to use HLC prices instead of close only
- Parse datetime column when loading enriched CSV
- Strip timezone from annotation timestamps
- Fix TA-Lib pattern names (CDL3WHITESOLDIERS, CDL3BLACKCROWS)
- Exclude programmatic label columns from training features
- Fix classification report to handle missing classes
- Update MLflow tracking to use localhost:5000
- Grant PostgreSQL permissions to ml_user

Pipeline now runs successfully end-to-end:
- Feature engineering: 2543 rows, 31 columns
- Annotation ingestion: 286 samples
- Training: 89.47% test accuracy with Random Forest

services/ml/training/evaluation.py

@@ -133,11 +133,28 @@ def generate_classification_report_text(
     Returns:
         Classification report as string
     """
+    # Get unique labels present in y_true and y_pred
+    present_labels = np.unique(np.concatenate([y_true, y_pred]))
+    
+    # If labels provided, use them as target names for the present labels
+    if labels is not None:
+        # If labels are strings, filter to only present ones
+        if isinstance(labels[0], str):
+            target_names = [label for label in labels if label in present_labels]
+        else:
+            # If labels are indices, map them
+            target_names = [labels[i] if i < len(labels) else str(i) 
+                           for i in present_labels]
+    else:
+        target_names = None
+    
     return classification_report(
         y_true,
         y_pred,
-        target_names=labels,
-        digits=4
+        labels=present_labels if labels is None or isinstance(labels[0], str) else None,
+        target_names=target_names,
+        digits=4,
+        zero_division=0
     )