Updated strategy for grouping GMOS Cals

[phirstgemini]

BIAS grouping needs to simply walk through dates and count files. Don't split nights. Can base counts on 1x1s for all combos, it won't be exact but close enough. When we hit the end of a group probably best to just count back from the end, there'll be some multiple use, but that's fine.

Imaging Twilight flats need to tie in to metric DB and aim for a total electrons per pixel count. HST uses 10^6.