Consider adding decision rules to fit beta_binomial and negativebinomial to overdispersed binomial and poisson data automatically

[beckyfisher]

At the moment bayesnec requires the user to examine the dispersion estimate and refit the model manually calling the appropriate family to handle the over dispersed case. Given the decision criteria are relatively clear in this case (dispersion significantly different from 1), it would make sense to automate re-fitting using the appropriate family.