Adds WAIC calculation

[atredennick]

I added another jags.sample call to calculate WIAC and the deviance of the model, as per https://sourceforge.net/p/mcmc-jags/discussion/610036/thread/8211df61/#ea5c. I was getting some warnings/errors when trying to compute DIC (e.g., pD was infitite). Some Googling suggested WAIC might be better when that happens. Likewise, WAIC seems to be preferred in general over DIC these days.

What do you think? We could also make an argument in the eoar function that let's the user decide whether to compute these or not since it is one more sampling step.

[jstudyvin]

@atredennick I like the idea of adding an argument (I say default it to FALSE, no information metric) and produce both the DIC and WAIC if the user wants them. It might also be worth throwing a warning when pD is infinite.

[atredennick]

Sounds good @jstudyvin, I agree. I'll add in the option and amend this pull request.

[atredennick]

I just added the argument for whether or not to compute WAIC and deviance. I also updated the documentation accordingly.