Neither covariant nor contravariant.

supertype[] is not a supertype of subtype[] if supertype and subtype alone are in that relationship, because the mutability of arrays means that the Liskov substitution principle doesn’t hold.

(These are all something you’ll probably find good explanations of on Wikipedia.)

That’s not realistic or “fair” - most Haskell projects will use a dozen or so extensions easily. GHC has been a platform for language experimentation for a long time; standardisation efforts keep on cropping up in annual surveys. (Eg, swapping in Text for String in base is long overdue, but it’s a hold over from days where FP pedagogy was seen as more important.)