Inaccuracy in moon phase calculation

[str4w]

There seems to be an inaccuracy of a few hours in the moon phase calculation. Unfortunately, this can cause the occurrence to switch from one day to the next fairly easily, which interferes with my use case.

Is an accuracy on the order of +/- 6 hours expected? If so, this is just a documentation problem. If not, I'll dig deeper.

Examples, comparing to the phase reported by the USNO
https://observablehq.com/@str4w/suncalc-test

+/- 15 minutes would probably be adequate for my use case.

[Artoria2e5]

The getMoonIllumination depends on the moon coords calculation, which might be the worst suspect here. https://www.aa.quae.nl/en/reken/hemelpositie.html#4 says:

The largest error in λ was 2.57 degrees (standard deviation 1.04 degrees), that in β 0.81 degrees (standard deviation 0.31 degrees), and that in 7645 km (standard deviation 3388 km).

These are quite a bit larger than the Sun approximations given above. Unfortunately I am too scatter-brained to do a differentiation and estimate how much time it takes for the moon to travel that much degrees to actually pin down the problem.

Anyways, see #102.