Add proper ESS/MCSE for discrete and binary and discrete variables

[bdeonovic]

[bdeonovic]

Hey @brian-j-smith just curious if you wanted this pulled in or if you needed more changes.

[brian-j-smith]

I have reservations about code that attempts to guess at variable types (continuous, discrete, or binary), similar to those expressed in #112. Such guesses could vary across different runs of a sampler for the same model, thus resulting in unreliable or unexpected results. Moreover, the user does not have control over the guessing criteria and cannot change the guessed type. That this issue has come up more than once now, leads me to believe a more global solution should be considered. In my opinion, incorporation of the variable types into the Chain structures, so that the types are set deterministically and changeable by the user, would be a more desirable solution.

[bdeonovic]

Thanks for the feedback. I agree with the global option of the user specifying which nodes are discrete in the model. I'll leave this pull here until some solution is decided on in the package