Add DataLoader module and introduce custom StopTraining exception

[SurbhiJainUSC]

Description

This PR introduces data_loader.py, a new module dedicated to encapsulating and streamlining the logic for loading and batching data for model training. This refactors data handling out of train.py, where it was previously imported by various trainers (e.g., sft_trainer, grpo_trainer, elastic_train). By centralizing this functionality, we significantly reduce the complexity of train.py and enhance overall code readability and maintainability. The DataLoader module now serves as a dedicated, modular, and efficient component for preparing training data.

Tests

E2E with train.py

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed.

[aireenmei]

LGTM! Thanks Surbhi for making this modularized and the unit tests!

[richjames0]

LGTM - thanks!