On the Impossibility of Learning the Missing Mass