[SFT] add warning if dataset's input_ids exceed max_length

[HERIUN]

What does this PR do?

setting default max_length=1024 may problem while long text dataset.

user have to know dataset's max_length and truncated with max_length.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a GitHub issue? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

Who can review?

@kashif @qgallouedec

[qgallouedec]

Thanks for the PR! I appreciate the intention behind adding visibility into truncation behavior. However, I'm not sure that raising a warning in this case is the best approach. Truncation can be a normal and intentional part of preprocessing, and emitting a warning during expected behavior can be noisy and misleading, especially in workflows where truncation is desired.

Instead, what might be more useful is a utility function (or method, I'm not exactly sure at this point) that prints or returns length statistics (e.g., quantiles) across the dataset. This would give users a clearer view of the lengths distribution and help them to choose the max length.