Improve the generation of QA pairs on tables

[ceberam]

The current version supports Q&A generation on tabular data. With the [ChunkingDocSerializer](https://github.com/docling-project/docling-core/blob/main/docling_core/transforms/chunker/hierarchical_chunker.py#L175 we can leverage tables from chunks in markdown format and fine tuned the LLM prompts.