Fairseq S2T: Fast Speech-to-Text Modeling with Fairseq