On the feasibility of a three-point bending setup for the validation of (fatigue) damage models for thin composite laminates