flake.nix: cleanup, drop flake-utils, stop shadowing pkgs

[faukah]

Changes done:

• Format using the official nixpkgs formatter instead of the outdated one.
• Stop shadowing pkgs.
• Drop flake-utils, instead opting for iterating over systems using pure Nix functions.
• Fix removing of ocamlformat properly:

nativeBuildInputs = lib.remove pkgs.ocamlformat (testNativeBuildInputs pkgs);

Removes pkgs.ocamlformat from testNativeBuildInputs, while testNativeBuildInputs uses the ocamlformat versio from .ocamlformat. If these ever mismatch, this breaks.

[Alizter]

Thanks for the PR. Can we split the first two commits off into a separate PR?

[spikespaz]

@Alizter Why does it matter if you avoid a merge commit? They're already separate changes, to show the cause (changing the formatter) and the effect (after formatting).

Edit: Though, they should be reversed.

[Alizter]

@spikespaz Its easier for reviewers when PRs are more focused. I asked for the first two commits to be split off because it is uncontroversial and is just a formatting change. The remaining commits will require more attention so it doesn't make sense to tie the first two down. Not sure what you mean about avoiding a merge commit however.

[Alizter]

Does this matter? I think we really wanted to shadow pkgs here but we can't so we introduced pkgs', however I think its quite confusing to mix using the two.

[faukah]

The mkShell in pkgs and pkgs' are the same, so it makes no difference; I'd argue that this is better though because everyone knows pkgs.mkShell. For pkgs'.mkShell you'd have to look at pkgs' first to be 100% sure mkShell is what you think it is.

It's wise to always keep the scope as minimal as possible for stuff like this.

[Alizter]

I can't tell if bumping the flake lockfile caused the issues, but we best do that in a separate PR. There might be unrelated changes that we will have to make and it will cloud this PR.