DocVXQA: Context-Aware Visual Explanations for Document Question Answering