GitHub doesn't understand stacked diff workflow very well.
gh is a tool that knows how to translate stacked diffs
into GitHub pull requests, in such a way that GitHub's
UI understands.

- Need a unique identifier for all patches.  Where to get this
  from?

    - Identifiers assigned by GitHub are right out, because
      we want to edit the identifier into the commit message
      we send, but we can't get an identifier until we actually
      perform a GitHub UI action.

    - Personal identifier cached on machine?  This doesn't work
      if you want to use the tool from multiple machines.

    - UUID?  That works, if a bit wordy.

    - Miniature stateful endpoint which keeps track of these for you?

    - How about: username + endpoint key + number?



1. OK, you have a diff stack.

2. Assign each diff a UUID, if you haven't already

3. Create a pull request, and durably associate it with
   the commit.  (Amend the commit message with the
   pull request)
