ML-DSA: update benchmarks pqclean

[franziskuskiefer]

This updates the pqclean benchmarks to use ml-dsa rather than dilithium.

[jschneider-bensch]

Thanks, that makes sense!
The criterion ones are not updated, though. Maybe this is a point where we can decide if we want to keep them? For me, they are still very inconsistent from run to run (large run-to-run diffs in signing, mainly) and for this reason I don't look to them at all, I find. So I would be fine retiring them. But you might get some use out of them still?

[franziskuskiefer]

Good point. I totally forgot about those. I updated them just in case we want to use them again...

Review re-requested