Fix/1370 efficientad validation and pretraining images

[jpcbertoldo]

Description

Fixes issues A and C in #1370

A: make it possible to split the train set to make a validation set.

C: make it possible to point to a directory with the images that should be used to the penalty term in the student loss of efficientad.

Known issue: the vocabulary in the classes for EfficientAD say "imagenette" everywhere, while it's more generally the image set used for the pretraining (could be something else). While I set the option name to pretraining_images_dir, i think it'd be better to align things.

Changes

Describe the changes you made

• Bug fix (non-breaking change which fixes an issue)
• Refactor (non-breaking change which refactors the code base)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update

Checklist

Ensure that you followed the following

• I have added a summary of my changes to the CHANGELOG (not for minor changes, docs and tests).
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas (If applicable)
• I have made corresponding changes to the documentation (If applicable)
• I have added tests that prove my fix is effective or that my feature works (If applicable)

[jpcbertoldo]

I additionally adressed an issue raise in this discussion #1368

The option pad_maps was removed to make the interface simpler. The behavior adopted is equivalent to padding XOR pad_maps (either the convolutions or the inference score maps are padded).

[Seanny123]

Would it make sense to make this parameter optional? That way, if it is None you can still do the automatic download of the Imagenet dataset.