feat: save automatic grid snapshot before manual edits for GT comparison

- build-grid now saves the automatic OCR result as ground_truth.auto_grid_snapshot
- mark-ground-truth includes a correction_diff comparing auto vs corrected
- New endpoint GET /correction-diff returns detailed diff with per-col_type
  accuracy breakdown (english, german, ipa, etc.)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

klausur-service/backend/grid_editor_api.py

@@ -31,6 +31,7 @@ from ocr_pipeline_session_store import (
     get_session_image,
     update_session_db,
 )
+from ocr_pipeline_regression import _build_reference_snapshot
 
 logger = logging.getLogger(__name__)
 
@@ -2814,8 +2815,22 @@ async def build_grid(session_id: str):
     except ValueError as e:
         raise HTTPException(status_code=400, detail=str(e))
 
+    # Save automatic grid snapshot for later comparison with manual corrections
+    wr = session.get("word_result") or {}
+    engine = wr.get("ocr_engine", "")
+    if engine in ("kombi", "rapid_kombi"):
+        auto_pipeline = "kombi"
+    elif engine == "paddle_direct":
+        auto_pipeline = "paddle-direct"
+    else:
+        auto_pipeline = "pipeline"
+    auto_snapshot = _build_reference_snapshot(result, pipeline=auto_pipeline)
+
+    gt = session.get("ground_truth") or {}
+    gt["auto_grid_snapshot"] = auto_snapshot
+
     # Persist to DB and advance current_step to 11 (reconstruction complete)
-    await update_session_db(session_id, grid_editor_result=result, current_step=11)
+    await update_session_db(session_id, grid_editor_result=result, ground_truth=gt, current_step=11)
 
     logger.info(
         "build-grid session %s: %d zones, %d cols, %d rows, %d cells, "