predict_and_eval: set→list to ensure IDs are sorted like lines

calamari_ocr/ocr/evaluator.py

@@ -257,5 +257,5 @@ def evaluate(self, *, gt_data: Dict[str, str], pred_data: Dict[str, str]):
         )
 
         res = Evaluator.evaluate_single_list(out, True)
-        res["ids"] = gt_ids
+        res["ids"] = list(gt_ids)
         return res