[feat] `response_format` differentiation

Additionally, for **docling**, I think it would be good to add something like a `response_format` parameter. Then:

- **default** – returns text, as it does now,
- **object** – returns the Docling object,
- **meta_data** – returns document metadata (such as the title inferred by the LLM and other details),
- **bbox** – returns bounding boxes (an object containing bounding box data), which will work with **easyOCR**.

The issue refers to the existing docling extraction path and easyOCR integration but names no files or tests. Start by tracing how docling currently produces the default text response; done means the response_format options consistently return text, the Docling object, metadata, or bounding-box data as specified.

Written by the indexing model from the issue text.