feat(streaming): enable per-dataset batch-sizes in CombinedStreamingDataset

[MagellaX]

Closes #327.
What does this PR do?

New capability – different batch-sizes per stream
CombinedStreamingDataset.set_batch_size(...) now accepts either
a single int (old behaviour), or
a Sequence[int] with one entry per wrapped StreamingDataset.
_CombinedDatasetIterator (for batching_method="per_stream") keeps yielding from the
selected stream until its own batch-size quota is reached, then switches to a new stream.
Implementation highlights
Core logic lives in:
src/litdata/utilities/base.py – validation & propagation of a Sequence[int].
src/litdata/streaming/combined.py – dataset-specific limit calculation and switching.
Single-int paths are untouched → 100 % backward-compatible.

Tests
Added test_combined_dataset_per_dataset_batch_size in tests/streaming/test_combined.py.
Parametrised with multiple batch-size lists; asserts that no stream emits more samples than its allotment.
All existing tests still pass (pytest -q).

Why is this useful?
Projects combining datasets with very different tensor sizes can now maximise GPU utilisation by
using a smaller batch on the “large” dataset and a larger one on the “small” dataset, without
resorting to gradient-accumulation work-arounds.

[deependujha]

Hi @MagellaX

Thanks for contributing to LitData! 🙌
Could you please take a look at the failing checks and fix the issues? Let me know if you need any help.

[tchaton]

Hey @MagellaX. Would you mind to fix the failing tests ?

[codecov]

Codecov Report

Attention: Patch coverage is 92.10526% with 3 lines in your changes missing coverage. Please review.

Project coverage is 83%. Comparing base (fc59c8a) to head (ad95aca).
Report is 1 commits behind head on main.

Additional details and impacted files

@@         Coverage Diff         @@
##           main   #635   +/-   ##
===================================
  Coverage    83%    83%           
===================================
  Files        49     49           
  Lines      6785   6812   +27     
===================================
+ Hits       5662   5688   +26     
- Misses     1123   1124    +1     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[bhimrazy]

Hi @MagellaX,
I've merged the latest changes from main to bring this branch up to date.
Whenever you get a chance, please pull the changes into your local branch before making any new commits to avoid conflicts.

Also, I'm reviewing the PR now 😊.

[bhimrazy]

Looking good so far! Added a few thoughts and queries.

[MagellaX]

Hey @tchaton @bhimrazy @lantiga @deependujha

Can you merge this PR! Everything is okay now!!