fix: make --resume in batch pipeline actually skip completed work

Previously --resume re-queried every gradient in every row, overwriting
existing values — an interrupted run could not be resumed cheaply.

- bicorder_query.py: add --resume flag; skip gradients whose cells already
  have values, and report how many were skipped
- bicorder_batch.py: pass --resume through to query; skip fully-complete
  rows before invoking the query script; report partial rows
- bicorder_batch.py: import row/config helpers from bicorder_query instead
  of calling undefined names (would have crashed on --resume)

analysis/scripts/bicorder_query.py

@@ -54,6 +54,16 @@ def get_protocol_by_row(csv_path, row_number):
     return None
 
 
+def get_row_values(csv_path, row_number):
+    """Get all existing values for a row (1-indexed) as a dict of column -> value."""
+    with open(csv_path, 'r', encoding='utf-8') as f:
+        reader = csv.DictReader(f)
+        for i, row in enumerate(reader, start=1):
+            if i == row_number:
+                return row
+    return None
+
+
 def generate_gradient_prompt(protocol_descriptor, protocol_description, gradient):
     """Generate a prompt for a single gradient evaluation."""
     return f"""Analyze this protocol: "{protocol_descriptor}"
@@ -153,6 +163,8 @@ Example usage:
                         default='../bicorder.json',
                         help='Path to bicorder.json (default: ../bicorder.json)')
     parser.add_argument('-m', '--model', help='LLM model to use')
+    parser.add_argument('--resume', action='store_true',
+                        help='Skip gradients that already have values in the CSV')
     parser.add_argument('--dry-run', action='store_true',
                         help='Show prompts without calling LLM or updating CSV')
 
@@ -177,6 +189,15 @@ Example usage:
     bicorder_data = load_bicorder_config(args.bicorder)
     gradients = extract_gradients(bicorder_data)
 
+    # Load existing values for this row (for resume mode)
+    existing_values = get_row_values(args.csv_path, args.row_number) or {}
+
+    # Count existing values for reporting
+    if args.resume:
+        already_filled = sum(1 for g in gradients if existing_values.get(g['column_name'], '').strip())
+        if already_filled:
+            print(f"Resume: {already_filled}/{len(gradients)} gradients already have values")
+
     if args.dry_run:
         print(f"DRY RUN: Row {args.row_number}, {len(gradients)} gradients")
         print(f"Protocol: {protocol['descriptor']}\n")
@@ -188,6 +209,11 @@ Example usage:
     for i, gradient in enumerate(gradients, 1):
         gradient_short = gradient['column_name'].replace('_', ' ')
 
+        # In resume mode, skip gradients that already have a value
+        if args.resume and existing_values.get(gradient['column_name'], '').strip():
+            print(f"[{i}/{len(gradients)}] {gradient_short}: SKIP (already has value)")
+            continue
+
         if not args.dry_run:
             print(f"[{i}/{len(gradients)}] Querying: {gradient_short}...", flush=True)