Data loading

[cmadrid1]

Add a page on preprocessing and specialized classes that build custom batches for training.

Generally, we need to have equal amount of data for different domains and manipulating the data is typically done in a custom data loader class. Most groups write this class from scratch every time but we should have documentation that can take a set of root files and build these custom batches.

The Weaver does this currently for PyTorch: https://github.com/hqucms/weaver