Adds more example tests to the CI
I'm opening this up as a starting point for discussion on exactly which examples the CI should tests. I've added me preference, but welcome to suggestions. A few points
- Testing marginalisation is slow because it does a precalculation.
- Testing open data examples are currently slow because they download the data. We could do something to see the testing flag and just generate fake data, but then is it a complete test?
@colm.talbot @moritz.huebner @plasky thoughts?