Upcast FSDP2 parameters only if requires_grad

[ojh31]

What does this PR do?

Fixes an unnecessary upcast of frozen parameters, particularly in the case of LoRA finetuning

Fixes #3844

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?

Fully-Sharded Data Parallism: @SunMarc @zach-huggingface

[SunMarc]

Thanks for fixing this ! Just a nit

[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.