A breakthrough on the Navier-Stokes equations could be worth a $1 million prize, but a mathematician has serious questions about how OpenAI got there so fast.