feature: add support for the geocentric Earth radius

[msheby]

Use WGS84 ellipsoid for latitude-dependent Earth radius in elevation adjustments

Add getGeocentricRadius() to AstronomicalCalculator that calculates
Earth's radius based on observer latitude using WGS84 ellipsoid.
ZmanimCalendar now uses this when elevation adjustments are enabled.

Earth's radius varies from 6378.137 km at equator to 6356.752 km at
poles. Using a fixed 6356.9 km introduced systematic errors of ~14 km
at mid-latitudes, translating to 1-2 seconds in sunset calculations.

Reference: WGS84 ellipsoid parameters (NGA Technical Report 8350.2)

[KosherJava]

@msheby , thank you very much for the PR. You may have noticed that there is a 3-0-modernization branch, and that is currently keeping me busy. I hope to look at this in a few weeks.

[msheby]

Oh, I didn't notice; would you like me to rebase my PR on that work?

[KosherJava]

@msheby ,
Again thanks for the PR. While it may take time for me to get to this, I have a few questions. Can you provide a source for this type of correction? Also, do you know if the SPA algorithm does something like this? There is a Java implementation at https://github.com/klausbrunner/solarpositioning/blob/master/src/main/java/net/e175/klaus/solarpositioning/SPA.java.
One more question. If someone is looking to calculate sunrise/set and factor in the elevation at the horizon, a simple way to accomplish this today would be to increase the sun's radius by the elevation of the horizon and subtract the elevation at the horizon from the viewer's elevation. To clarify
A viewer at an elevation of 5,000 meters who wants to know what time does the sun dip below the horizon if the elevation at the horizon is 1,000 meters?. The way to currently calculate this would be to increase the solar radius by 1000, and subtract 1000 from the viewer's elevation to get pretty close to the calculation that the user expects. With your approach this can't readily be accomplished since the radius can't be adjusted.

[msheby]

(First of all, I'm happy to get into the provinence of my contribution offline.)

With respect to your questions, yes, I'm familiar with the NREL paper. It explicitly implements the second edition of the Meeus book. Section 3.12 ("Calculate the topocentric sun right ascension") incorporates observer geographical latitude using the 1976 IAU values found on page 82 of Meeus. As my commit indicates, these values are the more recent WGS84 ones.

With respect to your last question, it might be cleaner to fold in rigorous altitude calculations than to alter the physical modeling to get what a user is truly interested in.