Parallel and Distributed Methods for Constrained Nonconvex Optimization—Part I: Theory