Fix S/N filtering in DM residual QA (b-arm DM residuals plot missing bottom)

The DM residuals task does some filtering on the lines and traces, which was added for Run 20 as part of PIPE2D-1600. The S/N cuts in particular had to be slightly different from the config param used for the adjustDetectorMap task and so a simple mean was used.

As part of the analysis for Run 20 it was noticed that in particular the b arm DMs were filtering out too many lines.

The code should be adjusted to use something smarter than the simple mean and/or the ability to disable the cuts (S/N, centroid, dispersion).