Updating to the current HEAD of the Tensorflow-Haskell bindings allows us to also update the dependencies, specifically proto-lens, and avoid having to retain their outdated versions.