Evolving mutation rate

[andyleejordan]

Part of issue #8 but most important

[andyleejordan]

I could also do what @Snewt is doing: since crossover takes care of nearby "neighbors" in the fitness landscape, the unit Gaussian mutation sequence is almost superfluous. What the mutation should instead do is, at say a 5 percent chance, change a gene to some other random valid value (not based on the current one), which will jump across hills quite a bit better.