Reducing skymodel confusion

[nicholebarry]

Some skymodel behaviour was very opaque, as pointed out by Shijiao. So, I set out to make it a little more obvious what is going on.

Previous functionality
When return_cal_visibilities is set, then the calibration visibilities are kept around after calibration. If a different set of visibilities need to be calculated for the subtraction model, only sources that are extra to the initial calibration sources list are calculated and added to the returned calibration visibilities. This assumes that the calibration sources are a subset of the model sources. While this reduces the number of calculations, it is not obvious from the user that setting return_cal_visiblities would cause this behaviour.

The skymodel that is saved, regardless of whether or not return_cal_visibilities is set, only includes model sources that are not "duplicates" of the calibration sources. There is no indication in the source list whether a source was used in the model. Assumptions include 1) that the model is a subset, and 2) that sources that are co-located to within 1/100th of a pixel are duplicates, regardless of brightness.

New functionality
When return_cal_visibilities is set, then the calibration visibilities are kept around after calibration. If a different set of visibilities need to be calculated for the subtraction model, all will be calculated unless combine_skymodels is set. This gives flexibility to do the previous functionality (and be more efficient in your number of calculations) or do something more complicated where the assumptions do not apply.

When combine_skymodels is set, the code attempts to combine the skymodels as above. Extra model sources are marked with unsetting the include_calibration keyword (an already present keyword) for that source in the skymodel; duplicates are not included. If it seems as though the model is not a subset of calibration, then all sources are saved.

If combine_skymodels is not set, then code will save separate skymodels.

save_skymodels must be set for skymodels to be saved.

[bhazelton]

This looks like a really good improvement, much more predictable behavior. one question I have is how does this interact with the current "standard" run? I feel like it doesn't -- the standard run doesn't use a different set of sources for calibration vs subtraction. Mostly wondering who all needs to be informed of this change.