We should use some code coverage framework to run the R package's unit tests. That would guarantee that builds fail:
- if any tests fail
- if tests don't cover at least
x% of the code
Could use this approach: https://github.com/jameslamb/ci_tools/blob/master/r/unit_tests.R