A State-Space Model for Decoding Auditory Attentional Modulation from MEG in a Competing-Speaker Environment