ToDo's for making it public

[PeterStaar-IBM]

Now

• make a folder converters and move docling inside converters
  • Add the VLM converter
• make a toplevel folder visualisation and move all related code into it
  • constants.py in current docling folder
  • utils.py in current benchmark folder
• make the dataset columns uniform (add Modality in all of them)
• make the README "sexy"
• move all code from converters.utils in benchmark.utils
• move converters.utils to evaluators.teds
• rename to create_pdf_docling_converter, create_img_docling_converter and create_smol_docling_converter
• rename BenchMarkColumns.DOCLING_VERSION to BenchMarkColumns.CONVERTER_VERSION

Later

• make a BaseConverter class, which is a virtual class and has two methods (convert and get_version()). The children should be
  • PdfDoclingConverter
  • ImgDoclingConverter
  • DoclingTableFormer !!!!!
  • VLMDoclingConverter

[cau-git]

make a BaseConverter class

This will be already part of the next gen implementation of the prediction provider interface. We will have this feature once we ported all dataset creation to this, but it should better be an iteration after making the initial version public, since everything works for docling right now.