[mock_uss, monitorlib] Add EGM96 datum conversion

[BenjaminPelletier]

U-space requires USSPs to deliver messages with MSL altitude (using EGM96 or EGM2008 geoid as the datum) to authorised users while ASTM F3411-22a transmits altitude between USSPs using the WGS84 ellipsoid as the datum. In preparation to incorporate MSL altitude observations, this PR implements an algorithm to determine the EGM96 geoid offset from the WGS84 ellipsoid and uses it to compute MSL altitude in mock_uss.

The computation is performed as part of the CI (during the NetRID tests). To verify accuracy, I used the following code:

import matplotlib.pyplot as plt
import numpy as np
from s2sphere import LatLng
from monitoring.monitorlib.geo import egm96_geoid_offset

for p in (LatLng.from_degrees(34, -118), ):
    dh = egm96_geoid_offset(p)
    print(f"{p.lat().degrees}, {p.lng().degrees}: {dh:.2f}")

lats = np.arange(-90, 90.1, 0.25)
lngs = np.arange(0, 360, 0.25)
geoid = np.zeros((lats.size, lngs.size), dtype=float)
for i, lat in enumerate(lats):
    for j, lng in enumerate(lngs):
        geoid[i, j] = egm96_geoid_offset(LatLng.from_degrees(-lat, lng))
plt.imshow(geoid, cmap="hot", extent=(0, 360, -90, 90))
plt.xlabel("Longitude (degrees)")
plt.ylabel("Latitude (degrees)")
cbar = plt.colorbar()
cbar.set_label("EGM96 geoid height above WGS84 ellipsoid (meters)", rotation=270)
plt.savefig("egm96.png")

The absolute accuracy of one point was manually verified (34,-118 -> -34.16 meters, versus this calculator), and then the overall shape was verified visually using this output:

[barroco]

Looks good to me 👍 Could you please add the scale on the plot you attached to the PR to help readers evaluate the accuracy?

[barroco]

It is not very clear why you added a buffer to the latitude argument and not for longitude, could you please clarify and perhaps add it in comments?

[BenjaminPelletier]

I've switched to use enumerated latitudes and longitudes rather than counting them separately here.

[barroco]

I guess the previous response will clarify the latitude buffer here.

[BenjaminPelletier]

The latitude enumeration now has a comment that it's on the [-90, 90] interval. arange does not have an option for an inclusive endpoint, so the targeted endpoint must extend further than the desired endpoint, but not so far as to reach the next step. Adding one half-step works for this.

[barroco]

Could you please clarify in a comment why latitude needs to be inverted here?

[BenjaminPelletier]

Added

[BenjaminPelletier]

Could you please add the scale on the plot you attached to the PR to help readers evaluate the accuracy?

Done; comment updated.