Empirical assessment of prediction intervals in Cochrane meta‐analyses