Implement SGDP4_NEAR_SIMP propagation

[JonathanMaes]

This implements the formula for propagating satellite orbits below 200km until theoretical ground impact.
I adapted this with the C implementation of SGDP4 and compared with Heavensat for satellites which are about to decay naturally, and the results seem to match.

I was having the same issue as #80 when propagating orbits, so now the NotImplementedError: Mode "Near-space, simplified equations" not implemented has been implemented.
As a consequence, instead of the NotImplementedError, the intended error Exception('Satellite crashed at time %s', utc_time) can now be reached when using Orbital._sgdp4.propagate.

• Closes NotImplementedError: Mode "Near-space, simplified equations" not implemented #80
• Tests added
• Tests passed
• Passes flake8 pyorbital

[mraspaud]

Thanks a lot for this PR @JonathanMaes ! As you can see the tests are now failing, probably because of some slight difference in the aiaa tests. Could you check what this is about?

[JonathanMaes]

It seems to fail because the NotImplementedError is no longer raised, so line 115 in test_aiaa.py no longer just continues to the next test-TLE (in SGP4-VER.TLE).
The first test-TLE that now fails is # ARIANE 44L+ R/B # Lyddane bug at <= 70 min for atan2(),.
This "Lyddane bug" causes the results to be wrong when compared to the aiaa_results file.

I am not familiar with this bug, and some searching online did not enlighten me all that much. It seems to be a bug that exist(ed) in SGP4 due to the discontinuity in atan, and affects some low-inclination/low-eccentricity orbits, though I can't seem to find how (or if) it was solved.

[mraspaud]

I looked quickly at this, and from what I can see, the failing test is not on Arian 44, but on SL-6 R/B(2), second instance (starting with 1 22312U). Nothing seems to be mentioned there about the Lyddane bug, and the error is quite significant: res[0] is 306.10478453, pos[0] is 257.91353553688265, leading to an error of about 48km in the x direction (tolerance is 5mm...)

Skipping that example, the next one to crash is COSMOS 2405, with an error of 5km in the z direction.

I looked at the changes you made though, and compared it to the C++ and matlab versions found here https://celestrak.org/software/vallado-sw.php, and I can't see anything wrong. So I'm really dumbfounded right now...