apply_chat_template: consistent behaviour for return_assistant_tokens_mask=True return_tensors=True

[mrsndmn]

What does this PR do?

This PR fixes the simultaneous use of flags return_assistant_tokens_mask, return_tensors, in apply_chat_template method.

Is related to #28950 and corresponding PR #30650

Before fix

output = tokenizer.apply_chat_template(
    conversations,
    tokenize=True,
    padding=True,
    return_dict=True,
    return_assistant_tokens_mask=True,
    return_tensors='pt',
)

assert isinstance(output['attention_mask'], torch.Tensor)  # ok
assert isinstance(output['assistant_masks'], torch.Tensor) # not ok: type of `output['assistant_masks']` is `List`

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 or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• 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.

@yonigottesman @amyeroberts @Rocketknight1

[mrsndmn]

Hi! Take a look, please

[Rocketknight1]

Hi @mrsndmn, I'm sorry we missed this PR! @yonigottesman can you take a look? It's okay if you don't have time - just let me know and I can take it instead

[mrsndmn]

Hey @yonigottesman, just checking in — did you have a chance to look at this PR? No rush, but let me know if you need anything from me!

[ArthurZucker]

LGTM let's merge 🤗