Add touchstone benchmarking of core functionality

[seabbs]

it is useful to use {touchstone} to benchmark runtimes across PRs so we can make sure we are improving the package. We likely want to benchmark both R and stan code and make sure to include real-world usage of stan code (i.e. in a model) this would suggest we include a stan model either in the benchmarks or in the package to facilitate this (similarly to #35)