[QoL fix] [Image processing] Add warning on assumption of channel dim

[aliencaocao]

Sometimes the input image can have height or width 1 and in HWC format. One example is if you np.asarray(PIL.Image). The current function to infer channel dim assumes the CHW silently which can cause confusion to users as the image will be wrongly processed as a grayscale image, which will raise further, harder to interpret errors when running through models.

This PR adds a warning when this happens. It also set the channel dim to last if inputs are PIL.Image as they are always in HWC after np.asarray, avoiding ambiguity in this case.

The default behaviour of assuming first dim as the channel has not changed to prevent any breaking changes.

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?

@amyeroberts

[amyeroberts]

Thanks for adding this! Flagging the ambiguous case is a good idea, although I suspect a very rare case (it implies the image has height or width dimension of 1 or 3).

[amyeroberts]

Thanks for iterating!

Just some tweaks and should be ready to merge

[amyeroberts]

Thanks for adding this warning and iterating on this!