Do Small Schools Improve Performance in Large, Urban Districts? Causal Evidence from New York City