Remove RichProgressCallback from examples

[lewtun]

What does this PR do?

This PR removes the RichProgressCallback from the example scripts because it (a) leads to very messy logs in distributed training, (b) produces hard-to-trace bugs like those encountered in #2043 and (c) introduces a lot of unnecessary branching logic to our scripts just for aesthetic purposes. It has also led to headaches in other libs like accelerate as noted in the issues below:

• Disable rich traceback configuration by default accelerate#1630
• Accelerate 0.13.0 and 0.13.1 break tracebacks in Jupyter-Lab accelerate#765
• AutoModel.from_pretrained is messing with stack traces in jupyter transformers#26085

See also this internal thread where rich interfered with cloud training: https://huggingface.slack.com/archives/C04EX6W3QSY/p1720008806136549

Questions

• Does this count as a breaking change since we are technically changing the aesthetics of the CLI?
• Would a less aggressive change be preferred, where I simply set TRL_USE_RICH=False by default?

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? 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.
• 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.

[lewtun]

For some reason this was only included in the SFT trainer and no others - I removed it for consistency

[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.