What to do when local edits exist on a conflicted file which is remotely resolved

[meejah]

A scenario was described during development of "resolve conflicts" PR #725:

• the _local_update signal of the MagicFile state-machine goes from _conflicted -> _conflicted without doing anything
• should we queue a LocalSnapshot, or is the user editing the file to ultimately resolve the conflict?
• what if a remote party resolves the file meantime, and so we get an incoming update that is also a resolution?
• (Could immediately re-conflict with the contents of the local edit -- that is, have a LocalSnapshot ready to go, apply the remote resolve but then produce a new conflict (hang on, we can't "produce a conflict" -- the other side has generated their new version) -- so yeah if we produce a LocalSnapshot at this moment, we could ensure its parent is our version before the incoming resolve (makes sense) and thus this is seen as a conflict by all parties)

That last thing might be the most-correct. That is:

• continue to do nothing when we're conflicted and see a local change
• _if_we get an incoming resolve while conflicted AND we have changed local contents, then:
  • queue a LocalSnapshot (with our current latest version as the parent, not the incoming resolve)
  • produce a new conflict locally (preserve "our" contents, write the resolution into conflict-marker
  • when our snapshot is uploaded, all other participants will see this as a conflict (I think?? right?)
• ...maybe the user completes editing the file and deletes the conflict-markers before any incoming resolution appears; this means we just create our own conflict-resolution Snapshot