consider adding `summary` arg (like for `runMCMC`) for nested approx

[paciorek]

This would be an option for sampleLatents and sampleParams, allowing user to get quantiles and perhaps mean/sd automatically.

Additional idea here would be to provide summaries automatically for fixed effects (INLA does something like this) but that would require us to distinguish fixed and random.

Easy enough to add, but I am focused on release at the moment and don't see this as critical for that.

[paciorek]

For reference, this is what we return for the MCMC summary:

> names(out)
[1] "samples" "summary"
> out$summary
            Mean        Median   St.Dev.   95%CI_low 95%CI_upp
mu    0.01790587 -0.0004211433 0.2802189 -0.57259922 0.8087039
sigma 0.15724089  0.0208625087 0.2547156  0.01339469 0.9346548