Add Data.Point containing mpoint, spoint and spoint_

[code5hot]

After seeing a reddit thread where someone said you'd have to be mad to want to implement sum without using fold as a justification for keeping it as a member of Foldable I decided to provide a reasonable facility for it which I think justifies not having sum as a member of Foldable.

import Data.Point
import Data.Monoid
import Data.Tree

ipsum :: (InitialPoint p, Functor p, Num a) => p a -> a
ipsum = getSum . mpoint . fmap Sum

t = Node "hello " [Node "world " [Node "foo " [],Node "bar " []], Node "baz " [Node "quux " [], Node "jimbo" []]]
u = length <$> t
c = ipsum u

c is now equal to 34 and the implementation isn't equivalent to fold because the guaranteed associativity of the monoid operation means subtrees can be summed individually and the results summed which also allows for highly parallel summing where fold doesn't allow it.

[code5hot]

I have fixes for ghc-7.8 but comments still very welcome

[kozross]

@RyanGlScott Should we considering merging this? I'm happy to do cleanup on it, but I'm not sure if we want it or not.

[RyanGlScott]

It's tough to say whether this belong in semigroupoids or not. These classes are almost what you'd get if you split of the fold method from Foldable, but they also lack fold's requirement that they be right-associative. Presumably, this is done because there are specific Monoid instances that are designed to exploit parallelism (although I'm not sure what they are off the top of my head).

[kozross]

@RyanGlScott Yeah, that sounds out of scope. Should we close then?