Optimal zeros for model reduction of continuous-time systems