Parametrize test_benchmarks.py test by criterion benchmarks

[roypat]

Currently, test_benchmarks.py contains a single test that runs all criterion benchmarks. With increasing number of benchmarks, we thus increase the duration of this test, and need to adjust its timeout. Instead, we want to parameterize the test by the list of criterion benchmarks we have in the repository. This means introducing a fixture that yields the benchmarks listed by cargo bench --all -- --list individually. Then the test itself would only run the benchmark it is provided. This should also make it easier to see which benchmarks are failing.

Originally posted by @pb8o in #4830 (comment)

[MacOS]

If someone has not done it yet, I would take it over.

[bchalios]

Hey @MacOS, please start working on it if you want.

[cm-iwata]

@MacOS
CC: @bchalios

Are you still working on this?
If you are busy, may I handle it instead you?