1. When every issue has a relative position set, we don't need to
   perform any updates, or calculate the maximum position in the parent.
2. If we do need to calculate the maximum position in the parent, many
   parents (specifically, groups with lots of projects) leads to a slow
   query where only the index on issues.relative_position is used, not
   the index on issues.project_id. Adding the GROUP BY forces Postgres
   to use both indices.