Default Image Data Augmentation

[basedrhys]

Is your feature request related to a problem? Please describe.
A common technique when training a CNN is to perform data augmentation (crop/shear/rotate/etc.) to improve the model's generalisability. This currently doesn't exist in WDL4J

Describe the solution you'd like
Some reasonable default augmentation methods to apply to the image data during training.

Additional context
Fast.ai has some good examples