Testing the benchmarks

[adespawn]

With the changes added in #198, I've added some sanity checks to ensure that the benchmark finishes successfully.

We can extend this idea and add launching the benchmarks as part of CI (probably just with n=1) to ensure nothing is being broken. This would help with running benchmarks (as it may happen that a single benchmark failure will require the whole benchmark re-run) and increase the test coverage (ideally, all of the things done in benchmarks should be tested in integration tests, but there may be some gaps in coverage that I'm missing)