drop lib's self dep and pass data to each function

[Pacman99]

Making lib dependent on devos, was not the greatest idea since it prevents others from using the devos lib. And theres quite a few nice lib functions that people might like that they can't use right now: mkPkgs, mkPackages, mkHomeConfigurations.

This gives the specific data to each lib function that that function needs.

[blaggacao]

I think we should rather go the contrary route and only pass deps that are strictly necesary down to second and third order functions.

If a humongous, all-encompassing self needs to be passed through more than one hop, we should rather ask ourselves if it's time for refactoring and untangling the code, before loading it with more functionality.

I think revising/refactoring is in first order, before we can / should implement this feature. Otherwise, we risk in getting this wrong and I don't know the costs of cleanup vs the benefits of the feature.

[Pacman99]

all this PR is doing is reverting an old change. So with or without this, cleanup is necessary. But with this, we can get #218 merged, which makes lib a lot easier to work with, for me at least. And will allow out of tree mkflake usage.
I think the benefits far outweigh the slight harms.

[Pacman99]

I think this is the best way to do this, its hard to combine them all into one argument even though they are related. I chose this method as to not add more implementation logic to combine overlays.

[blaggacao]

LGTM, as I understand this PR is a cleanup to limit internal api surface! Nice!

[Pacman99]

Converted to draft, so we only merge once 0.9 is in

[Pacman99]

I think it might be good to include this in 0.9(contrary to what I said before), but its not necessary.

[nrdxp]

bors try

[bors]

try

Build succeeded:

• ci/hercules/evaluation
• ci/hercules/derivations

[blaggacao]

Superseded by #251