The feature, powered by Google's Nano Banana 2 AI model, had allowed users to generate photorealistic images grounded in Google Earth's satellite, aerial and 3D imagery by typing a text prompt for any location on the globe