Rewrap instead of forking

A good part of maintaining the Heptapod fork of hg-git would probably better be done upstream, and it is easier to have a clear separation of concerns.

In the developments for Heptapod 0.6, I've noticed that the most heavy-lifting we're doing to map topics, follow obsolescence etc could very well be done from an outside extension, let's call it heptapod.

Also, that would be the proper place to implement heptapod#73 (closed) and improve transactional properties (heptapod#92 (closed))