Add cross_validate helper

[snotskie]

A generic cross_validate helper should be straightforward to write:

1. Given an existing model and k, split unit ids into k groups
2. Run i in 1:k trainedmodels, being careful to preserve model config except for an added unitFilter and'ing on the existing one, holding out group i
3. Report stats summarizing the k models' stats in their embeddings

This should put a little heat under the problem that ENA models often don't cross-validate

[snotskie]

Maybe do this as an Example instead of a built-in helper?

[snotskie]

Use as a good example for https://github.com/snotskie/EpistemicNetworkAnalysis.jl/issues/13

[snotskie]

https://github.com/snotskie/EpistemicNetworkAnalysis.jl/commit/2d2af2f863e7602c0d45b5499b7de4140f60aa9d