Fix multi gpu loss sync condition, add doc and test

[techkang]

What does this PR do?

Multi-gpu loss sync is controlled by the average_tokens_across_devices argument. However, it has a tiny bug (it doesn't work when the model uses a user-defined loss function) and lacks unit tests.

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?

@muellerzr @ArthurZucker @SunMarc

[techkang]

Also fixed: #35078

[muellerzr]

Thanks for getting custom loss functions working as well under multi-GPU!

cc @SunMarc

[SunMarc]

SGTM !