[FEATURE] Run benchmarks in GHA #574
Description
We only build the benchmarks in CI, but we don't run them. This means that benchmarks can fail to run without us noticing, as shown for example in #541. Ideally, we would run the benchmarks in CI as well, so that this breakage does not go unnoticed.
One challenge is that the benchmarks can take a while to run, and that would further inflate running time for CI. Maybe there would be a way to run the benchmarks in a lightweight mode with smaller inputs, based on some command-line flags?