Properly sync data in rename_column_concurrently helper

The rename_column_concurrently helper can be used to rename a column without downtime. However, in some situations this helper won't properly maintain data consistency between the old and new column:

It creates a copy of the original column, and a trigger to update the new column whenever the old column is written to. But, during a deployment we have some servers running both old and new code. So old code is writing to old column, and that is copied to the new column, which is fine. But new code is writing to the new column, which is not being copied to the old column, so any servers still reading from the old column will not see those changes.