JOIST: A Joint Speech and Text Streaming Model for ASR