Barycentric -> Geocentric conversions when rephasing celestial coordinates

[bennahugo]

Although still well within the instrumental resolution there is a problem in using current ICRS celestial coordinates and rephasing far off the original zenith position in celestial coordinates. If we build larger TARTs this will become more of an issue, so I'm documenting this now.

The UVW bases are in geocentric frame, which means the l,m and n coordinates should also be in geocentric frame. However, we are using the ICRS coordinates in computing the cosines, which are in barycentric reference frame.

See discussion in codex-africanus#.

I don't think it really affects things that are in orbit of course, including the model we use to ultimately calibrate phases, but if we are tracking a distant object one day this will be an issue in the astrometry