Optimal Combination of Neural Temporal Envelope and Fine Structure Cues to Explain Speech Identification in Background Noise