Microsoft Releases Florence-2: A Novel Vision Foundation Model with a Unified, Prompt-based Representation for a Variety of Computer Vision and Vision-Language Tasks