Two new matrix-variate distributions with application in model-based clustering