feat(streaming): native datasets-5 episode batching and worker-split suppression

Allow datasets 5.x (pin >=4.7,<6; lockfile moves to 5.0.0) and use its
Arrow-native batch(by_column="episode_index") (huggingface/datasets#8194
sibling, #8172) for episode admission when available - one Arrow
accumulation per episode instead of one Python dict per row - with the
existing row loop as the 4.x fallback. A parity test asserts both paths
group identically.

Also fixes a latent worker bug this surfaced: `datasets` detects torch
DataLoader workers and re-splits its shards internally (_iter_pytorch),
on top of our explicit per-worker shard assignment. That second split
silently drops data whenever a per-worker stream has fewer internal
shards than there are workers (masked so far by single-file test
fixtures), and on datasets 5.0 it crashes by_column batching outright.
The worker context is now hidden from `datasets` while draining streams
we already partitioned (process-local patch, restored on exit).

The multi-shard shuffle buffer (huggingface/datasets#8194) is
intentionally NOT used: frame-level shuffling upstream of episode
grouping would fragment episodes and break delta windows. Its threaded
multi-source prefetch idea remains a follow-up for episode admission if
fetch timings warrant it.

Verified on both datasets 4.8.5 (fallback) and 5.0.0 (native): 27/27
streaming tests each; full datasets suite 469 passed under 5.0.0.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

tests/datasets/test_streaming_native.py

@@ -353,3 +353,28 @@ def test_fast_forward_resume_with_dataloader_workers(tmp_path, lerobot_dataset_f
     assert resumed == full[samples_consumed:], (
         "fast-forward resume with DataLoader workers did not continue at the exact sample"
     )
+
+
+def test_episode_grouping_native_and_fallback_agree(tmp_path, lerobot_dataset_factory, monkeypatch):
+    """The datasets>=5 batch(by_column=...) path must group episodes identically to the row loop."""
+    import lerobot.datasets.streaming_dataset as sd
+
+    repo_id = f"{DUMMY_REPO_ID}-grouping"
+    _make_local_dataset(lerobot_dataset_factory, tmp_path / "ds", repo_id, total_episodes=5, total_frames=100)
+    ds = StreamingLeRobotDataset(repo_id=repo_id, root=tmp_path / "ds", shuffle=False, max_num_shards=1)
+
+    def episode_signature(use_native):
+        monkeypatch.setattr(sd, "_HAS_BATCH_BY_COLUMN", use_native)
+        return [
+            (ep_idx, [int(row["index"]) for row in rows])
+            for ep_idx, rows in ds._iter_shard_episodes(ds.hf_dataset)
+        ]
+
+    fallback = episode_signature(False)
+    assert len(fallback) == 5
+    if not sd._HAS_BATCH_BY_COLUMN and "by_column" not in str(
+        type(ds.hf_dataset).batch.__doc__ or ""
+    ):  # datasets < 5: only the fallback path exists
+        return
+    native = episode_signature(True)
+    assert native == fallback