moon_phase_angle slow for large time arrays

[kvyh]

I found that when I tried to use MoonIlluminationConstraint when scheduling, it was very slow. When running a profiler on the code, I found that for an array of 90 times:
moon_coord = get_moon(...).transform_to(altaz_frame) took 26 seconds
sun_coord = get_sun(t).transform_to(altaz_frame) took 7 seconds
sun_coord.separation(moon_coord) took 9 seconds.

The information is cached (so it only has to do this once for all targets), but it takes nearly .5 seconds per time which adds up for large time ranges and fine time resolutions. Is there a way to speed this up while retaining precision? Otherwise, for finer time resolutions it may be beneficial to have an interpolation method so that only a fraction of the times need to be calculated.

[kvyh]

#213 solves part of the problem, but MoonIlluminationConstraint still calls Observer.moon_altaz which didn't get changed in it.

[StuartLittlefair]

This is fixed in #286 and #285.

Benchmarking on #286 I find that for 100 times:

moon_coord = get_moon(...).transform_to(altaz_frame) takes 461 ms
illumination = np.array(moon_illumination(times)) takes 344 ms

the total call to evaluate a MoonIlluminationConstraint for 100 times and a scalar target is 780ms. For two targets the vectorisation in the above PRs means it still only takes 811 ms.