Bring back `ShowHO` and `EqHO` as we had before about generics-mrsop

[VictorCMiraldo]

Related to #42 #41

[serras]

I am not at all surprised. I've had so many cases where I start using All and then it falls short at some point. One idea I have player with is to make versions of NP and NS parametric on a constraint, like:

data NP (c :: k -> Constraint) (f :: k -> Type) (xs :: [k]) where
  (:*) :: c x => f x -> NP c f xs -> NP c f (x ': xs)
  Nil :: NP c f '[]

but then you might be pushing QuantifiedConstraints to their limits.

[VictorCMiraldo]

Yeah, I start to dislike All very much; it really hinders the usage of the library with overcompliated typeclass engineering and most of the times, it doesn't add that much safety.

This is one of the points we make on our mrsop paper and came back to agreeing with it! :)

I like having the same NS and NP as sop-core, because it allows us to port mrsop functionality back and forth easily, I think the best design really is to just go back to using our versions of EqHO and ShowHO with mrsop but using NS and NP from sop-core; this would also drop the need for -XQuantifiedConstraints.

I will sleep on it though; if I feel inspired and sure this is the option; I'll implement on monday and release another version of mrsop; it is slowly converging to something more stable! :)