Tackling the infinite likelihood problem when fitting mixtures of shifted asymmetric Laplace distributions Preprints