DPLL(T) with Exhaustive Theory Propagation and Its Application to Difference Logic