Display 5 worst predictions at the end of evaluation

Based on !339 (merged)

Closes #245 (closed)

dan/ocr/evaluate.py

@@ -89,7 +89,14 @@ def eval_nerval(
         if not (ground_truths and predictions):
             continue
 
-        scores = evaluate(ground_truths, predictions, threshold)
+        scores = {
+            key: {
+                k: round(value * 100, 2) if k in ["P", "R", "F1"] else value
+                for k, value in values.items()
+            }
+            for key, values in evaluate(ground_truths, predictions, threshold).items()
+        }
+
         print(f"\n##### {split_name}\n")
         print_results(scores)