I think the scenario shadeless was describing is that the author of the PR shouldn't be able to merge but anyone else with write access could. It would be my ideal scenario as well. (You could of course bypass this by doing it in Git if you really needed to.)
Gotcha, so if three users have write access then when a PR is opened by one of them, one of the other 2 users has to approve it?
If your scenario only has two users (you and one other person) then this flow can already be accomplished (the two are both reviewers, and you set approval required by "everyone" -- so you approve your own PR and then the other user has to also). If you've got more than two users, that's something we'll have to add but it sounds like it's worth doing!
