Fairness and Incentives in a Multi-Task Principal-Agent Model