If we want to add more parallelism later, then it is worth discussing
future developments of the API now, so we don't put ourselves into a
corner. If the answer is no, it is clear we are only parallelizing tests or
parallelizing modules, but never both. And if we want to do both in the
future, we
I think the answer is clearly NO currently.
1. Being able to run test cases in parallel is a big win already.
I assume for most resource-light code bases (like state-less libraries),
they can just upgrade to `async: per_test` for all their tests, without
worrying about resource utilization.
2.
