[feat] research minerU + pdf-extract-api

Let's check if it's somehow useful for us to add an extraction integration with the https://github.com/opendatalab/MinerU/tree/master/demo

By the way I really like their pipelines architecture which makes multistage extraction and parsing more scalable