SDG and RAG today consume the input data directly, use docling and further format it to prepare the data prior to kicking off their respective workflows. 

We want to pull out the data ingestion and pre-processing functions into the core library so both SDG and RAG can consume it (and, eventually other libraries like Eval etc).

A part of this work will be done in 1.4 where the SDG library (where this is housed today) will split it into three different APIs: pre-processing, SDG and post-processing. It will continue to live in the SDG library, and we'll start transitioning it to Core in phases.