Skewed Component Distributions or Transformations? Modelling Skewness in Cluster Analysis