fix regression: current docling version fails for specific pdf while earlier version works

[samhita-alla]

Bug

I'm trying to convert a PDF to Markdown format. Docling version 2.2.0 works as expected, but the latest version, 2.15.1, does not. The newer version only outputs image tags, whereas version 2.2.0 correctly includes both text and image tags in the output.

Steps to reproduce

from docling.document_converter import DocumentConverter

source = "..."  
converter = DocumentConverter()
result = converter.convert(source)
print(result.document.export_to_markdown())

Here's the PDF that didn't work for me: https://github.com/NVIDIA-AI-Blueprints/pdf-to-podcast/blob/main/samples/investorpres-main.pdf

Docling version

2.15.1

Python version

3.12.2

[cau-git]

@samhita-alla I am reproducing this with Docling 2.17.0 and confirm there is most content detected as picture only. It will need some deeper analysis on the layout model.