Restructure data pipeline

Establish an interface for computing and visualizing arbitrary statistics on labeled data, similar to how markov_stats.py creates a graph with counts of observed actions. This interface should read from a library of functions defining statistics.