[pagged-attention] fix off-by-1 error in pagged attention generation

[kashif]

What does this PR do?

we were adding the token first and then doing a check, leading to the number of tokens being 1 more than themax_new_tokens

This fix reorders the logic to:

1. Check completion conditions (EOS token or max length) before adding the token
2. Only add the token if we haven't exceeded the limit (with special handling for EOS tokens)
3. Terminate the request appropriately

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.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[ArthurZucker]

Thanks for updating! 🤗