Don't barf on 'HsSpliceTy'

[harpocrates]

This handles HsSpliceTys by replacing them with what they expand to. IIUC everything that is happening, renameHsSpliceTy should not be able to fail for the inputs we feed it from GHC.

This fixes #574.

---

Troublesome snippet from secret-sharing that was causing the error (FYI, this is primeField):

-- | A finite prime field. All computations are performed in this field.
newtype FField = FField { number :: $(primeField $ fromIntegral 1021) }
  deriving(Show,Read,Ord,Eq,Num,Fractional,Generic,Typeable,FiniteField)

Snippet of Haddock produced after this PR

[alexbiehl]

Nice! Thanks Alec (btw. could you open your next PRs to the GHC-8.4 branch? It will become the default branch once ghc-8.4.1 is released)

[philderbeast]

Is there a chance that this small fix could be added to the GHC-8.2 branch too please? I'm currently depending on uom-plugin that fails to build on the GHC-8.4 branch and I'm hitting the HsSpliceTy problem using stackage lts-11.8 when building docs.

adamgundry/uom-plugin#43
GlideAngle/flare-timing#31