Performance-Based Incentives in a Dynamic Principal Agent Model