Refactor scaler to util

[muellerzr]

What does this PR do?

This PR adds a new func to utils.modeling in the form of get_grad_scaler.

There's a few times when we need to repeat some of the logic for grad scalers for backward compatibility etc (see #3132) and also easier to have typos (https://github.com/huggingface/accelerate/pull/3116/files) so instead keeps one reference and refactors to its own func.

I debated on making this a class, but a functional getter similar to the one for mixed precision seemed ideal

Fixes # (issue)

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.

@BenjaminBossan @SunMarc

[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.

[BenjaminBossan]

Generally LGTM, thanks. Just one small comment and the question whether the function should also be used in benchmarks/fp8/ms_amp/ddp.py and benchmarks/fp8/ms_amp/non_distributed.py.

[SunMarc]

LGTM ! +1 on @BenjaminBossan comment