Move cmdstan installation to GitHub Action.

.github/workflows/main.yml

@@ -31,4 +31,8 @@ jobs:
             any::rcmdcheck
             cmdstanr=url::https://stan-dev.r-universe.dev/src/contrib/cmdstanr_0.8.1.tar.gz
           needs: check
+      - name: Install cmdstan
+        run: |
+          Rscript -e 'cmdstanr::check_cmdstan_toolchain(fix = TRUE)'
+          Rscript -e 'cmdstanr::install_cmdstan(${{ env.cmdstanVersion }})'
       - uses: r-lib/actions/check-r-package@v2

vignettes/getting_started.Rmd

@@ -17,12 +17,6 @@ knitr::opts_chunk$set(
 )
 ```
 
-```{r, include = FALSE}
-# Make sure cmdstanr is all set up. But we don't need to show the reader this.
-cmdstanr::check_cmdstan_toolchain(fix = TRUE)
-cmdstanr::install_cmdstan()
-```
-
 `gptoolsStan` is a minimal package to publish Stan code for efficient Gaussian process inference. The package can be used with the [`cmdstanr`](https://mc-stan.org/cmdstanr/) interface for Stan in R. Unfortunately, [`Rstan`](https://mc-stan.org/rstan/) is not supported because it [does not provide an option to specify include paths](https://discourse.mc-stan.org/t/specifying-include-paths-in-rstan/32182/2). If you're already familiar with `cmdstanr`, dive in below. If not, have a look at the [getting started guide](https://mc-stan.org/cmdstanr/articles/cmdstanr.html) for the `cmdstanr` interface.
 
 This vignette demonstrates the package by sampling from a simple Gaussian process using Fourier methods (see the accompanying publication ["Scalable Gaussian Process Inference with Stan"](https://arxiv.org/abs/2301.08836) for background on the approach). Here is the model definition in Stan.