Tweak defaults for quantized-typed FP8 TE weights

[muellerzr]

What does this PR do?

Per the NVIDIA team, we should be using HYBRID for training, E4M3 during inference. As a result, tweaks the default to be HYBRID

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

[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]

In general, apart from the line length, LGTM.

Do the benchmarks and low_precision_training.md also need to be adjusted?

[muellerzr]

Benchmarks actually used HYBRID from the start (since I was looking at nvidia's examples directly) so those are fine :)