Test faster svd / pca variants

[gabrielodom]

The package currently uses the "mysvd()" function to extract vanilla PCs for the supervised PCA approach. Test alternative pca / svd algorithms (and even an if {} else {} setup for the cases when the pathway dimension is greater than the sample size).

[gabrielodom]

I've tested a few svd / pca variants. I think I'll use the RSpectra package. The svds and eigs functions are almost twice as fast as the corpcor::fast.svd function (which is incidentally slightly slower than base::svd). The only rub is that this technique (I think) precludes the possibility of specifying the percent of variation to preserve.

[gabrielodom]

After speaking with Prof. Chen, we will take a hard look at the time difference between using base::svd and RSpectra::svds for a real pathway set. This is because coprcor::fast.svd is marginally--but not significantly--slower than the base SVD function. If the time difference between RSpectra and base is practically significant, then we will use the SVD function from this package; otherwise, we'll cut out another package call and use base instead (this is my preference).