Parallelize ECSW

[eparish1]

I was looking at the code, and I think we can make a parallel version fairly easily. The pinv is the main challenge I see, but we can do something similar to method of snapshots where we use pinv(A) = inv(A^T A) A. We can do A^T A in a distributed manner, which gives a small n_s x n_s matrix. We can then do then inv locally. Other than this, I think most operations are just parallel mat-mat and mat-vec as well as min/max.

We will hold on this until #126 #70 are further along.