Dataset subset caching

[edhenry]

The nature of the windowing performed for the creation of the distributed representations of NetFlow data call for a caching of the subset data that was used to produce the model. This will allow for easier reproducibility in the future.

One should be able to specify a data set directory within the config.ini file which will then attempt to save out the data set as it is generated for training of the model.