[feature] Augmentation function over a collection of images stored in a Numpy array.

## 🚀 Feature

I would like to apply a single augmentation function over a collection of images stored in a Numpy array. I tried passing an `(K,L,L)` array of `K` black and white images of size `LxL` to `aug_np_wrapper`, but it doesn't seem to work.

## Motivation

I'm working on a project that would require transform all of the images in a single dataset, e.g., Fasion MNIST.

## Pitch

I would like to be able to apply a transformation over a given axis for a collection of images.

## Alternatives

The only alternative I can consider at the moment is using a for-loop, but this will be quite inefficient for I would like to obtain.

Start by inspecting the aug_np_wrapper entry point and how it currently handles image-shaped NumPy arrays. Check whether existing augmentation behavior defines an axis convention for collections of images. Done means a single transformation can process an (K,L,L) collection as requested, with behavior that is documented or covered by the project's existing validation approach.

Written by the indexing model from the issue text.