A remark on normal forms and the upside-down I-method for periodic NLS: growth of higher Sobolev norms