[Feature]: and support PaddlePaddle OCR in the backend

Describe the proposed feature

There is a new really strong ocr model which should be supported out of thr box: https://github.com/PaddlePaddle/PaddleOCR