Reorder split names

Reorder split names in the YAML metadata to be aligned with the order in the loading script.
See: https://github.com/huggingface/datasets/issues/5258

README.md

@@ -39,12 +39,12 @@ dataset_info:
     - name: answer_start
       dtype: int32
   splits:
-  - name: test
-    num_bytes: 3865097
-    num_examples: 3643
   - name: train
     num_bytes: 7712523
     num_examples: 7182
+  - name: test
+    num_bytes: 3865097
+    num_examples: 3643
   - name: validation
     num_bytes: 1072731
     num_examples: 1000