Revisiting reliability with human and machine learning raters under scoring design and rater configuration in the many-facet Rasch model.