Allow for within-chain parallelization using reduce_sum()

[BrandonEdwards]

At least part of the model files (namely, the count ~ poisson_log() or count ~ neg_binomial_2()) can take advantage of Stan's reduce_sum() feature to allow for within-chain parallelization. I am unsure by how much this would speed up performance because it is only a very small part of the sampling, but could be worth a shot.

I am happy to take care of this because I am about to implement it on my version of bbsBayes2, and so I can also add the code into here.