Leveraging Whisper Embeddings for Audio-based Lyrics Matching