rework dataset function and class structure

tests/test_RLBootstrapDataset.py

@@ -5,6 +5,7 @@ from pathlib import Path
 import tempfile, gzip
 
 from server.dataset.RLBootstrapDataset import RLBootstrapDataset
+from server.dataset.DatasetIO import DatasetIO
 
 class TestRLBootstrapDataset(unittest.TestCase):
   def test_count_jsonl_rows_reads_gzip_dataset(self):
@@ -15,17 +16,15 @@ class TestRLBootstrapDataset(unittest.TestCase):
         handle.write("\n")
         handle.write('{"turn":2}\n')
 
-      self.assertEqual(RLBootstrapDataset.count_jsonl_rows(dataset_path), 2)
-      self.assertEqual(RLBootstrapDataset.count_jsonl_rows(Path(tmp) / "base.jsonl"), 2)
+      self.assertEqual(DatasetIO.count_jsonl_rows(dataset_path), 2)
+      self.assertEqual(DatasetIO.count_jsonl_rows(Path(tmp) / "base.jsonl"), 2)
 
   def test_rotate_and_gzip_if_size_reached_rotates_jsonl(self):
     with tempfile.TemporaryDirectory() as tmp:
       path = Path(tmp) / "rl_bootstrap.jsonl"
       path.write_text("x" * 200, encoding="utf-8")
 
-      rotated = RLBootstrapDataset.rotate_and_gzip_if_size_reached(
-        path, max_bytes=50
-      )
+      rotated = DatasetIO.rotate_and_gzip_if_size_reached(path, max_bytes=50)
 
       self.assertTrue(rotated)
       self.assertFalse(path.exists())