Speed up computations for larger graphs

[mnarayan]

For whatever reason QUIC seems to be quite slow ones n_features > 200.
Need to dig into to how to improve this.

Possible solutions include regression based estimators that wrap around high dimensional regression #58 and #74 could come in handy. Both of these can be combined with spark compatibility helps for parallelization at the level of nodes or pathways.