Text this: Interpretable Embeddings for Next Point-of-Interest Recommendation via Large Language Model Question–Answering