Could there be further refactoring to the `seq` typing rule
Seq has a special cased typing rule. It's been greatly simplified in aab3c6d1 . But there is still duplication, with a different rule applying when
seq is used in infix position (unchanged in that commit), and when used in prefix position.
Could there be a way to reduce code duplication here? I find it really hard to convince myself that both cases do the same thing.
Though, really, given levity polymorphism, does
seq need a special typing rule at all anymore?