Use readout exp time for dark cal

Related to PR 145 from common; we need to make and correct for dark current by considering the exposure time per readout (an FPA can be made of multiple readouts). Conceptually this is straightforward: we just normalize all arrays by the number of frames in a single FPA prior to dark correction. In other words, all data are now normalized so that the values represent the counts per-readout.
This means that the thing that needs to match between Dark Cal objects and other tasks is the readout exposure time, NOT the total FPA exposure time.

The diff here is big, but that’s only because dark correction happens in a lot of places!