more algorithms for .torsion_basis() over {finite, number, ""} fields

[yyyyx4]

This patch adds an implementation of the method for general fields (including number fields), as well as adding new algorithms for the method over finite fields. With this, we address the comment

        # TODO: In many cases this is not the fastest algorithm.
        # Alternatives include factoring division polynomials and
        # random sampling (like PARI's ellgroup, but with a milder
        # termination condition). We should implement these too
        # and figure out when to use which.

in src/sage/schemes/elliptic_curves/ell_finite_field.py.

The purpose of the "random" algorithm is to address the issue that the current implementation often wastes a lot of time (sometimes exponential) on computing the entire structure of the group of rational points, when in reality only a small subgroup is needed. Here's an example that showcases this problem:

sage: F.<t> = GF((2^127-1, 4))
sage: E = EllipticCurve(F, [1,0])
sage: %time _ = E.torsion_basis(2^128, algorithm='random')
sage: %time _ = E.torsion_basis(2^128, algorithm='structure')

Output:

CPU times: user 133 ms, sys: 0 ns, total: 133 ms
Wall time: 134 ms
CPU times: user 6.72 s, sys: 0 ns, total: 6.72 s
Wall time: 6.73 s

[github-actions]

Documentation preview for this PR (built with commit fc3d03a; changes) is ready! 🎉
This preview will update shortly after each push to this PR.