Wingman incorrectly ignores skolems when dealing with GADT evidence

[isovector]

Wingman uses dataConCannotMatch from GHC to check when it should filter out GADT data cons. But dataConCannotMatch only checks if two type indices certainly can't match. But if one of them is a skolem, GHC doesn't know this, and thinks it's merely a type variable. In which case, it could match, but we know better.

[isovector]

dataConCannotMatch calls typesCantMatch, which explicitly is glomming all tyvars as univars:

https://hackage.haskell.org/package/ghc-8.10.2/docs/src/Unify.html#typesCantMatch

Easy fix --- just need to reuse the skolem-aware unifier from wingman.