Cache intermediate NeuroMMSig results

[cthoyt]

EpiCom reloaded runs pretty slow, and one of the reasons is that it re-calculates the betweenness centralities and hubs for each subgraph EACH time. This only needs to be calculated globally once for each subgraph, and can save a lot of time, especially since these graphs are small and use the (slow) exact algorithm.