fix: rewrite GRPO data pipeline for offline record-level access

- process_list_field returns List[List[int]] preserving per-response boundaries
- GRPODataset rewritten to record-level __getitem__ (no windowing/stride)
- grpo_collate_fn pads variable-length responses into [B, G, R] tensors
- JsonlStore detects nested List[List[int]] and stores List[Tensor] per record
- Store._normalize skips nested-list keys from cumsum bookkeeping
- Pipeline._flush handles nested lists without cross-record flattening
- Export grpo_collate_fn from astrai.dataset
- 6 new GRPO tests + 2 updated builder tests, 114 total pass

astrai/dataset/__init__.py

@@ -1,6 +1,7 @@
 from astrai.dataset.dataset import (
     BaseDataset,
     DatasetFactory,
+    grpo_collate_fn,
 )
 from astrai.dataset.sampler import ResumableDistributedSampler
 from astrai.dataset.storage import (
@@ -21,6 +22,7 @@ from astrai.serialization import (
 __all__ = [
     "BaseDataset",
     "DatasetFactory",
+    "grpo_collate_fn",
     "Store",
     "StoreFactory",
     "H5Store",