Use ddem maps from CTL run for NO_FILT, NO_GAP runs

Ideally, we should save the non-filtered, non-gap filled ddem difference for CTL runs, and use that for `NO_FILT` and `NO_GAP` runs, and we should not be computing the linear regression each time with the same set of co-registered and bias corrected DEMs.