[Feature request] Support external image URL for Multi Modal Models / Vision LLMs

1. download the image 2. load the image 3. run inference on image 🎉 4. profit 🤑 This is especially useful if you're running ollama on a server and you can't just drag and drop an image _Ideally_ ``` $ ollama run minicpm-v --verbose >>> https://farmhouseguide.com/wp-content/uploads/2021/08/group-of-llama-ee220513.jpg Added image './group-of-llama-ee220513.jpg' The image shows a group of lamas gathered around a water source in an outdoor, mountainous landscape. There are six animals visible: four white llamas with thick woolly coats and two reddish-brown guanacos or vicuñas. The setting appears to be high-altitude terrain with sparse vegetation and rocky ground. ```