Allow LabelMaker to return data split into test/train sets or test/train/validation

[martham93]

Hello!
@wronk and I think it will be important for future label-maker users to allow either a test/train set split or a test/train/validate split.

To address this, we can open a PR that modifies package.py and have it take in two additional lists as parameters.

• One list will be a list of strings to name the data subsets it will default to ['train', 'test'] , but people could pass ['train', 'test', 'validate']
• Other list a list of floats that must sum to 1, and will default to [.8, .2], but people could pass [.7, .2, .1].

cc @drewbo