Replace ab_biprod_corec_eta with grp_prod_corec_natural

[jdchristensen]

Since we generalized ab_biprod_corec to grp_prod_corec, we should also generalize ab_biprod_corec_eta to grp_prod. In the process, I renamed it from _eta to _natural. It's a definitional equality that isn't used anywhere, but is probably worth keeping as a form of documentation.

I had to switch grp_prod_corec to use wildcat notation, and I switched a couple of nearby things as well.

[Alizter]

Switching GroupHomomorphism to $-> will introduce a strictly larger universe in those expressions, but as we've said before we don't care when algebraic universes land. (Which by the way are still being worked on, just optimizing the new algorithm is a bit slow).