Comparing the Lasso predictor-selection and regression method with classical approaches of precipitation bias adjustment in decadal climate predictions