Note [Solving Wanted QCs from Given QCs]

When we are about to solve a Wanted quantified constraint, and there is a
Given quantified constraint with the same type, we should directly solve the
Wanted from the Given (instead of building an implication).

Not only is this more direct and efficient, sometimes it is also /necessary/.
Consider:

  f :: forall a k. (Eq a, forall x. Eq x => Eq k x) => a -> blah
  {-# SPECIALISE (f :: forall k. (forall x. Eq x => Eq k x) => Int -> blah #