Fix: ignore precompute_ref_log_probs when use_liger_loss=True

[ginkyenglee]

What does this PR do?

Problem: When use_liger_loss=True and precompute_ref_log_probs=True, the trainer still computes and attaches unused reference log probability columns (ref_chosen_logps, ref_rejected_logps). This results in unnecessary overhead.

Solution: Force-disable args.precompute_ref_log_probs (set to False) when use_liger_loss=True, and emit a warning log to inform the user.

Impact: Limited to the DPOTrainer initialization path for the Liger loss setting. The change is backward compatible and does not affect other training modes.

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?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[kashif]

shall we add this in the dpo config's post init?

[ginkyenglee]

shall we add this in the dpo config's post init?

I moved this to DPOConfig.__post_init__ in the latest push.