Robust Multi-Objective Feedback Design with Linear Guaranteed-Cost Bounds