Data, embedding, and index of MassiveDS by "Scaling Retrieval-Based Language Models with a Trillion-Token Datastore"