Let iterable dataset shard have a length if implemented

[muellerzr]

What does this PR do?

This PR mimics transformers implementation for the IterableDatasetShard by bringing over the ability for it to have a defined len and allow for epoch-based training rather than number of batches.

Fixes # (issue)

fixes #2023

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 @LysandreJik

[BenjaminBossan]

Thanks for the fix, looks good.

Tbh I don't understand the underlying logic, it seems to be subtle, but since it's a 1:1 copy from transformers, I guess it's fine leaving this as is.

[LysandreJik]

Thanks @muellerzr!