Add support for Deepseek OCR

### What would you like to be added?

Deepseek OCR support via Ollama https://www.ollama.com/library/deepseek-ocr

### How will it help?

Way more accurate OCR in theory bounding box support could also be added for screenshots as the model can do that too overall it's just way more flexible at the cost of way bigger hardware requirements

### Extra info

_No response_