Survey of Vision-Language-Action Models for Embodied Manipulation