Reduce GH actions testing costs

[johnjasa]

We got an unofficial request through the grapevine to scale back the amount of GH actions tests we're running.
This is probably because we have a swath of tests across different Python versions that run for every PR and commit pushed to the PRs.

We could reduce the minutes used by having the full testing suite only run on non-draft PRs or those marked "ready for review," per @RHammond2's suggestion.
Me thinking this through, we could also have just a subset of the tests run (e.g. just unit tests for one version of Python) for PRs not marked ready, or something like that.

Another obvious but more time-consuming way to reduce our testing computational cost is to reduce the scope and depth of the tests we run.
We could examine the cases that are run, cut out the unnecessary ones or simplify them, and see where that gets us.

Definitely worth more thought.

[RHammond2]

As discussed in previous standups, another course of action to take is to pare back the examples in the repo, or design specific tests around the use case. Similarly, it would be ideal if instead of rerunning the same case many times, we use fixtures to store the commonly reused run cases and test those more thoroughly.