The Optimal Size of an Epistemic Congress