save_pretrained: use tqdm when saving checkpoint shards from offloaded params

[kallewoof]

Even at 9B, doing a save with partially CPU offloaded tensors can take up to 30 seconds. Doing a 70B model can take 5-10 minutes depending on how much VRAM you have around. This adds a progress bar to give the user some visual feedback whenever there are offloaded params.

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?

[amyeroberts]

Thanks for adding! LGTM as we do the same for loading checkpoint shards