GH actions now run with the default settings, which do not allow to scale once more complex tests/devices are added to the test suites.

To have fast and scalable CI testing, self-hosted GH runners might be used to better parallelize and complete all the tests in a reasonable amount of time.

At the moment all the tests are executed with a comprehensive make target here: https://github.com/SymbiFlow/fpga-interchange-tests/blob/0dcffe28969010e288256d770b7414e0cf7e615a/.github/workflows/ci.yml#L49-L51

To have a matrix-style CI, we may need to add more specific make targets, for instance all-counter-tests or all-arty-tests. The choice must be based on the wanted granularity for each job.