Text this: PairAlign: A Framework for Sequence Tokenization via Self-Alignment with Applications to Audio Tokenization