An LLM pre-training dataset containing only public domain and openly licensed text