A better approach to centering

[alexpghayes]

Karl from lab meeting:

(I - n^{-1} 1 1')A
that is row centering
A ~ UDV'
(I - n^{-1} 1 1') A ~ (I - n^{-1} 1 1') UDV'
so... we can svd(A)
then column center U

Then cholesky?? to reorthogonalize. (QR?)

[alexpghayes]

cc @karlrohe @MuzheZeng I am very on board with implementing this change but would prefer that someone else sort out the mathematical details and send me a quick writeup, rather than figuring them out myself.

[MuzheZeng]

@alexpghayes I emailed you.