Don't vendor Ranged-sets?

[Ericson2314]

Right now, https://hackage.haskell.org/package/Ranged-sets depends on big deps like HUnit and QuickCheck. But I have a PR PaulJohnson/Ranged-sets#4 so there is a flag to disable the arbitrary instances and it just depends on base. At that point, I don't see why we just delete the vendored code and use that instead?

[andreasabel]

A reason could be that atm, Alex depends only on libraries that ship with GHC (https://wiki.haskell.org/Core_Libraries_Committee#Core_Libraries). This might be a property worth preserving. (But I don't have insight into the GHC build process, for instance.)

[Ericson2314]

Yeah I won't pull the trigger on this until I bring up on ghc-devs. Maybe we could have a flag to either use or not use the vendored part as a stop-gap to track divergence too.