Fix default attention mask of generate in MoshiForConditionalGeneration

[cyan-channel-io]

What does this PR do?

Fixes #36160, which is caused by the absence of default attention mask initialization when attention_mask is not given.

_prepare_attention_mask_for_generation is mostly borrowed from MusicGen implementation.

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.

@ylacombe, @eustlb

[SeungyounShin]

Hi everyone,

I just wanted to bring this PR to your attention, as it addresses a critical issue where generation in MoshiForConditionalGeneration doesn't work at all due to the absence of a default attention mask.

I understand you're all busy, but if you could find some time to review this, it would be greatly appreciated. This fix was primarily adapted from the MusicGen implementation to ensure consistency and reliability.

Thanks in advance for your time and consideration!

cc. @ArthurZucker, @amyeroberts, @younesbelkada, @ylacombe, @eustlb

[Rocketknight1]

cc @eustlb and @gante for generation

[gante]

Thank you for identifying the issue and fixing it 🤗

[gante]

@eustlb I've confirmed that this fixes the root issue, including the script in this comment, so I'm taking the liberty of merging it 🤗

@cyan-channel-io @SeungyounShin Thank you for making the library better for everyone 🙌