This caches the rendered commit partials in MRs, using a
helper to calculate the required cache key. In the MR context
we don't need to cover all aspects of the view.