Code review comment for lp://qastaging/~alan-griffiths/mir/doodle-in-TimeoutFrameDroppingPolicy-code

Well given that alarm cancel is back to the original semantics (like in 0.3.0) then swap_unblocked should attempt to cancel alarm first so that if it returns false, then it means the callback handler is running or about to run, so pending swaps should not be touched or the alarm rescheduled (since it will be done by the callback handler)