Extensions to model-based clustering for mixed-type data : a new model framework, variable selection, and outlier detection