Option for neutral simulations without stop codons

[ebrintn]

From @jasondk:

"Okay. I think that's reasonable if modelling pseudogene evolution, but if we are doing neutral simulations of protein coding genes, the suitable model still probably assigns stop codons zero (or close to zero) fitness. Typically codon models simple exclude stop codons from the model entirely, reducing the state space from 64x64 to 61x61 (for standard gen code). I think there are good reasons to do it either way...
It could be useful to have an option to run neutral model with stop codons assigned zero fitnesss. But i think what you've implemented is more general"