Summary:
For some asinine reason, we were suppressing warnings when
deriving associated type family instances with `DeriveAnyClass`. That seems
like a bad idea. Let's not do that.

Along the way, I noticed that the error contexts associated with these
newly emitted warnings were less than ideal, so I did some minor refactoring
to improve the story there.

Fixes #14094

Test Plan: ./validate

Reviewers: bgamari, austin

Subscribers: rwbarton, thomie

GHC Trac Issues: #14094

Differential Revision: https://phabricator.haskell.org/D3828