Myna: Masking-Based Contrastive Learning of Musical Representations