Make the augmentation script less clunky

[NotJoeMartinez]

The augmentation script has the following issues:

• It requires that you change the root_dir from the training to testing script manually meaning you have to run it twice if you want to augment the testing dataset
• It uses a shell subprocess to copy datasets calls a function that does the same thing from scripts/separate_datasets.py

[NotJoeMartinez]

Fixed with a42a5e8