do weighted LS for fpc_wsvd

[fabian-s]

tf/R/tfb-fpc-utils.R

Line 40 in 429afd9

scores <- t(qr.coef(qr(efunctions), t(data_wc) / sqrt(w))) #!!

should prob. also weight by variance of "unexplained rest" from truncated eigenfunctions to avoid overfitting in high-variance regions of domain...