Sobolev Training of End-to-End Optimization Proxies