Generalise foldS worker to have a different accumulator type

[Ben Lippmeier]

With

foldS :: (Elt a, V.Unbox a)
      => M.IOVector a   -- ^ vector to write elements into
      -> (Int# -> a)    -- ^ function to get an element from the given index
      -> (a -> a -> a)  -- ^ binary associative combination function
      -> a              -- ^ starting value (typically an identity)
      -> Int#           -- ^ inner dimension (length to fold over)
      -> IO ()

the worker could really have type (a -> b -> b) in the sequential case.