Benchmarking Nix CIs

It's been said (including by us) that Nix is a great way to prevent vendor lock-in on CI, since at their core any Nix CI is doing the same thing. This also means it should be possible to accurately benchmark these CIs.

So that's what we did. We picked some well-known open source projects that have Nix already set up, forked them, and ran various different CIs on the same commits.

You can see the results, and a lot more details, here.