Test all supported versions in test matrix
Why
It seems historically we've opted to test the least and greatest supported versions.
It seems like the tests aren't very expensive, so we should start testing all the supported versions to ensure we're getting the best coverage possible at a reasonable cost.
What
- create an MR to add all the supported versions
- create a separate MR to add more context to the README. This chore seems to come up regularly, and it should be easy to see what our testing philosophy is from the README.