Type cast before normalize layoutlmv2

[amyeroberts]

What does this PR do?

Modifies the feature extractor call to make the inputs to BatchFeature always numpy arrays. This ensures the type of the returned pixel_values matches those requested with return_tensors. The conversion to numpy arrays happens before normalization to ensure consistent rescaling is done on the inputs.

This solution means the return_tensors=None default can stay and that the behaviour is as expected for any combination of the do_xxx flag values.

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?

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.