Save the outputs of shuffle

Martin Reinecke points out that the outputs of shuffle need to be saved somewhere, and this should be added to the datamodel. We'd want to know at least the position (ra, dec), proper motion/parallax, magnitude, colour, which AG, and AG (x, y). The metadata will need to know the design epoch and telescope elevation (maybe 2020 and zenith).

One option would be to add an extra HDU to the pfsDesign file.

We also need to worry about refraction, so we'll need to update the (x, y) at the time that we generate the pfsConfig file; this suggests that the pfsDesign proposal is a good way to go.