Estimating mixtures of 6D diffusion tensor distributions with "Magic DIAMOND": are b-plates better than b-spheres?