Properly vectorize propagators

[astrojuanlu]

At the moment, the "vectorization" step for most propagators happens outside of numba, harming performance. For example:

poliastro/src/poliastro/twobody/propagation.py

Lines 146 to 151 in e57d8c4

	results = [farnocchia_fast(k, r0, v0, tof) for tof in tofs]
	# TODO: Rewrite to avoid iterating twice
	return (
	[result[0] for result in results] * u.km,
	[result[1] for result in results] * u.km / u.s,
	)

This is related to #921 and perhaps should be tackled afterwards.

cc @Josvth

[astrojuanlu]

Prospective contributors: consider looking at #921 first, then we can work on this.

[astrojuanlu]

@andrea-carballo expressed interest in working on this! The steps should be:

1. Move test_propagate_with_coe to a newly created tests/tests_core/test_propagation.py
2. Move the functions in core/propagation/__init__.py to core/propagation/vallado.py, core/propagation/mikkola.py... as done with core/propagation/farnocchia.py and adapt tests and imports accordingly
3. Write farnocchia_coe_tofs, mikkola_coe_tofs, markley_coe_tofs, pimienta_coe_tofs, gooding_coe_tofs, danby_coe_tofs doing this:

@jit
def propagator_tofs(k, ..., tofs, ...):
    nus = np.zeros_like(tofs)
    for tof in tofs:  # pseudocode
        # call corresponding propagator with single tof (time of flight) and store nu
    return nus

This doesn't fully fix the issue, but it's a first step.

[astrojuanlu]

Postponing this to the next release.

[astrojuanlu]

Blocking this on #921.

[astrojuanlu]

I think #921 will need a significant refactor, so I'm postponing this issue. At least will try to tackle #921 for the next release.

[astrojuanlu]

#921 is about to be fixed. We can deal with this for the next release.

[astrojuanlu]

Now propagators have a propagate_many method that still is iterating on the Python side rather than on the numba side:

poliastro/src/poliastro/twobody/propagation/farnocchia.py

Lines 59 to 65 in 3abced1

	results = np.array(
	[farnocchia_rv_fast(k, *rv0, tof) for tof in tofs.to_value(u.s)]
	)
	return (
	results[:, 0] << u.km,
	results[:, 1] << (u.km / u.s),
	)