Facial Expression Recognition Using a Hybrid ViT-CNN Aggregator