Migrate away from Eq1 / Ord1 / Show1

[Bodigrim]

Closes #27.

Tested with

packages: .
packages: dhall-1.42.1
packages: aeson-2.2.1.0
packages: hnix-0.17.0
packages: sexp-grammar-2.3.4.2
packages: recursion-schemes-5.2.2.5
packages: quickcheck-instances-0.3.30
packages: liquidhaskell-boot-0.9.8.1

allow-newer: hnix:aeson, hnix:free, hnix:bytestring, hnix-store-core:algebraic-graphs

I'm on aarch64 so unfortunately cannot test locally with old GHCs.

Really old GHC <= 7.6 fail in CI. If we must retain compatibility with them, I can revert changes with regards to Typeable1.

[Bodigrim]

@phadej I got access to an x86_64 machine and tested the patch with GHC 8.4 and 8.6. No breakage in downstream packages detected.

[Bodigrim]

@phadej anything else you would like me to check here?

[phadej]

Everything looks fine.

I'm waiting for GHC-9.10 release; as this is still a breaking change, I'd like to make a 9.10 compatibility work first, and only then make a major release.

[phadej]

Now thinking more about it. I'm not happy about this change. It exposes too much of implementation detail to the surface. Ord1 f is asking for more, but doesn't tell anyone on the outside that Fix is used internally.

I'm not particularly happy with this. So I'm sorry, but we'll need to rethink this change.

Changing Fix interface is one option, but I don't like that either, as then Fix and Mu/Nu parts will look different.

[phadej]

One option, which is not breaking, and may fix part of #27 is to use forall a. Ord a => Ord (f a) implied by Ord1 f in the implementation, when such implication exists (i.e. more recent base).

[Bodigrim]

One option, which is not breaking, and may fix part of #27 is to use forall a. Ord a => Ord (f a) implied by Ord1 f in the implementation, when such implication exists (i.e. more recent base).

Works for me. Do you want it only for Mu and Nu or for Fix also? Shall I update the PR?

[Bodigrim]

Closing in favor of #29.