The entire staging process is now done locally with no interaction
with the repo until the staging branches get pushed to.
Removed the backoff / visibility check, as we expect the git repo to
be a "better" reference than the API, and we're now pushing there
directly.
Implemented rebase (-ish?) by hand, following the model of the one
built on the github API, so we have all the information we need *and*
still don't need a local working copy.
Had to update *several* tests due to formatting changes from using the
CLI git (maybe that could be avoided by feeding the commit message
through stdin but w/e).
Also added a dupe of an existing helper for the formatting changes
bit, will have to remove it once all's done.
If the stagings are going to be created locally (via a git working
copy rather than the github API), the mergebot part needs to have
access to the cache, so move the cache over. Also move the maintenance
cron.
In an extermely minor way, this prefigures the (hopeful) eventual
merging of the ~~planes~~ modules.
