A p-power neurodynamic approach to distributed nonconvex optimization