Expose integration settings

[seabbs]

We should consider exposing integration settings in both stan and R. For stan the only thing to expose is setting the relative tolerance. This could be nice for compute constrained use cases but the downside is we can't set defaults in stan easily so doing so could reduce readability. One option could be to overload just a few functions with defaults set but the problem with that is we then wouldn't be able to load those functions into R for testing.

[seabbs]

Currently we might not want to do this as any change to this setting causes a large increase in model failures (see #34)

[seabbs]

Due to #64 this is now an issue for exposing ODE settings.