Text this: ViE-Take: A Vision-Driven Multi-Modal Dataset for Exploring the Emotional Landscape in Takeover Safety of Autonomous Driving