End-to-end test for static numbers

[zsusswein]

Currently we have end-to-end tests, but these are more "smoke tests" checking that the code runs all the way through. They run the sampler for only a couple of iterations and write files, so the results aren't quantitatively rigorous.

Ideally we would have a more heavyweight end-to-end test that ran only in CI. This test would run on simulated data (our gostic_toy_rt package data) and several chains and and draw a couple thousand samples. We could use this test to check for unexpected changes in the underlying model.

I think a decent check would be to run the model, check that everything was written as expected and the diagnostics are clean, and to plot the summaries vs the data and known simulated Rt. These plots could be snapshot tests through vdiffr::expect_doppelganger(). We could control the test only runs in CI with a testthat::skip_if and setting an env var in the CI environment.