Gradient descent, documentation, errors, context

Yup, you can now perform a sort of gradient "ascent" on your genomes as long as they implement the new GenomeGrad interface. If you use float types as your genes from the genes package all you need to do is implement the LenGrad and GetGeneGrad methods.

Also in v0.8:

• More expressive errors+Move common errors out to mu8 root level package
• Move FindCodependencies to mu8 (previously in genetic)
• Advance now takes a context, no more storing of contexts in structs
• Document all the things

Full Changelog: v0.7.1...v0.8.3