On testing the log-gamma distribution hypothesis by bootstrap