Trust-region problems with linear inequality constraints: exact SDP relaxation, global optimality and robust optimization