Improve speciation algorithm and interface

[HyperCodec]

I'm still not satisfied with the new implementation of the speciation feature.

First of all, I think it's bad to require Eq and Hash in the Species type. Perhaps I can fix this by using a BTreeMap, but then I need PartialOrd.

I also don't like the way the algorithm works. It seems naive, slow, and messy.

[HyperCodec]

I think we just remove Species altogether, and just have a simple method is_same_species where users can implement whether two genomes are compatible with each other.

[HyperCodec]

going with an implementation based on what was designed for NEAT, as I think an algorithm that is capable of preserving novel structures even in the complex NEAT setup would work well if applied broadly to the rest of genetic algorithms. not using innovation numbers and such tho.