Add script to run tests through combinations of `Sys.CPU_THREADS` and `Threads.nthreads()`

CI is slow.
Might be faster to also execute this script on CI instead of starting a bunch of short jobs, so things like installing packages and precompiling only happen once.

CI is spending a lot of time waiting for jobs to start, so because we aren't getting parallelism in this setup, we are paying that cost.