Rethink the pystrometry dependence

[j-faria]

If this package is only used for the astrometry phase plot, and introduces the pyarrow dependence, we need to consider possible alternatives.
For now I'll remove it from the dependencies and move the import inside the astrometry_phase_plot function.

[ThomasBaycroft]

Yes, I couldn't find any other way to get the plot working which requires knowing the parallax factors of any given star at all times in ra and dec and pystrometry can calculate this. I agree that figuring it out so that pystrometry is not needed would be good