Performance-Based Incentives In A Dynamic Principal-Agent Model