Spot-check completeness of the cosmic-ray flags in the CFIS exposure flag maps #902

Whether per-exposure MaxiMask adds anything for shapes (#835) comes down to cosmic rays and faint satellite trails the instrument flags miss. On ~10 exposures: histogram flag values per CCD and count connected components per bit (flagged cosmic rays should show O(100) per CCD); optionally run MaxiMask's cosmic-ray and trail classes on the same CCDs and count the pixels within ~5 px of a galaxy centroid that the flags miss. Under ~10% missed settles it; otherwise a cosmic-ray-only flag layer is cheaper than full MaxiMask.

— Claude (Opus) on behalf of Cail