Add comments

[ito4303]

Added explanations on the generation of posterior predictive distribution

[bob-carpenter]

I'm going to merge this, but I'm a bit curious, because the comments say:

// This code generates the posterior predictive distribution
// unconditioned on the observation

A posterior predictive distribution is by definition conditioned on some observations, it's $p(\tilde{y} \mid y)$ where $\tilde{y}$ is new data and $y$ is observed data.

[bob-carpenter]

P.S. Thanks for continuing to maintain this. We still use a lot of these models for bulk testing and people still use them as the basis for models they develop.