move hyperelliptic curve reconstruction to invariant theory

[mstreng]

As discussed in #22173, the genus-two hyperelliptic curve invariants are just binary sextic invariants. And the reconstruction is just binary sextic reconstruction. These can be applied more generally. Also, there will be too many things in the global namespace and too many classes if we continue in the old way.

• Move the internal workings of the invariants and reconstruction to the same place as where the quintic case (#26370, #25508) is implemented.

• Deprecate HyperellipticCurve_from_invariants and add its functionality to HyperellipticCurve (with keywords) in such a way that it is also ready for higher genus, move Mestre_conic from the global namespace to the invariant_theory module.

• Add method invariants to a general class of (hyperelliptic?) curves regardless of genus, which accepts a keyword to select invariants, and does Igusa-Clebsch invariants in case of genus 2.