Alignment performance improvement

[zairving]

Currently, when reducing data all three filters are aligned separately. This is not only computationally expensive, but can also cause issues if one filter has a poor SNR. It would therefore be better if only the "best" (i.e., highest SNR) filter was aligned, and then these alignments were reused for the other filters (accounting for pixel-scale differences and such).