Separate and Reconstruct: Asymmetric Encoder-Decoder for Speech Separation