support for Baidu's PaddleOCR (supports gpu, newer, and more actively developed compared to Tesseract)

In case this is still being maintained, it would be nice if it could also support PaddleOCR.

I have a working fork here with some added features (e.g. cropping, the ability to skip frames) that I wouldn't mind merging back into the original: https://github.com/oliverfei/videocr-PaddleOCR