Provide standard kwarg for model fitting random seed

[juntyr]

I am trying to make PCA fitting using xeofs.models.EOF reproducible by always selecting the randomized solver and passing a random_state kwarg to the solver. Unfortunately, not all solvers that xeofs uses internally, in this case, use a random_state parameter, one also uses seed, and neither accepts the other. At the moment, I need to loop over these possibilities, try each, and catch the type error if it was the wrong one. Could the model init or fit method take a random_state or seed parameter that is then passed to the correct kwarg for the chosen solver? Thanks for your help and the fantastic library!

[nicrie]

Thanks for bringing this up, sounds reasonable. I'll see if I can include it in the next release.

[juntyr]

Thank you so much for the quick fix!