Incorporate "Shintaro feedback"

[cjordan]

(A WIP branch shintaro-feedback contains some work for this that may be useful)

Calibration is best when the sky model matches what's in the data. Our sky models cannot have a priori information of the ionosphere in any observation, so our sky model will always be sub optimal due to the ionosphere. But, if/when peeling is working, it is possible to take the ionospheric constants for arbitrary sources and apply them to the sky model during DI calibration. This should yield better calibration solutions.

To be (hopefully) clearer, a workflow:

DI cal -> peel -> power spectra

is not as good as

DI cal -> peel -> DI cal with better model -> peel with better-calibrated data -> power spectra.

My approach in the code is to have the ionospheric constants read in via a json file, and modelling is done differently for any sources in common with the sky model and the iono constants. The modelling code in particular needs a lot of work; the phase centre will have to be changed when modelling sources with iono constants, because the constants are only correct when the expected source position is the phase centre. This will slow down the modelling of these sources. Any sources without iono constants (or iono constants of 0) can be modelled normally.