Fix cmdstanr setup

[VisruthSK]

Also cleanup docs and conform to lf

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 790f29e is merged into version-0.2.0:

• ✔️Statistical_Rethinking_with_brms_ggplot2_and_the_tidyverse: 1.31s -> 1.31s [-0.56%, +0.77%]
• ✔️ggplot2: 190ms -> 189ms [-2.03%, +0.95%]
• ✔️loo: 820ms -> 817ms [-1.26%, +0.57%]
• ✔️loo_knitr: 580ms -> 579ms [-0.93%, +0.63%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

Plots