Implement latent distance and stochastic block model in new framework

[slinderman]

The old versions of pyglm relied on graphistician to support different network models. To reduce dependencies, I've copied the most important network models into pyglm. I still need to copy over the latent distance and SBM code though.