Fix extra_reference_date logic for single reference unwrapping

[scottstanie]

The current logic correctly relabelled the interferograms across the reference-date change; however it used the same reference phase (by using convert_pl_to_ifg) and didn't apply the new reference phase.

The other change is to fix the unclear logic about when to apply the reference date, if it's supplied: the only times we ignore during wrapped_phase are if we're using spurt (i.e. is_using_spurt) or if we've requested nearest-3/nearest-n interferograms (is_using_short_baseline_ifgs). The logic looks different because the old one was subtly wrong, but just didn't come up in practice.