Table omission in conversion from Docling Document to markdown

### Bug

When converting a document from PDF to Markdown, **the entirety of the second table in the document is omitted**. Upon investigation, this issue persists regardless of the PDF backend and OCR engine used, and was only slightly fixed by using the VLM setting (although this caused a different set of inaccuracies).

When reviewing the Docling Document preview and JSON with Docling-Serve's UI, it appears that the table and its content are captured correctly. The ref to the table appears as a child of the body, and the content in the table is correct. However, the Markdown format was again missing the table (now converted with Docling-Serve rather than Docling in a Jupyter Notebook).




### Steps to reproduce
To reproduce the bug exactly as I did, run [the notebook in this repository](https://github.com/Ryfernandes/docling-table-omission/blob/main/multiple_conversion_settings.ipynb) (runs conversion for 15 different PDF backend and OCR engine configurations, can remove to just convert with default settings).

Alternatively, download the (publicly accessible, no sensitive content) file below and convert it to markdown with any non-VLM Docling pipeline.

[02_BofA_CoreChecking_en_ADA.pdf](https://github.com/user-attachments/files/20658899/02_BofA_CoreChecking_en_ADA.pdf)

### Docling version
Docling version: 2.33.0
Docling Core version: 2.31.1
Docling IBM Models version: 3.4.3
Docling Parse version: 4.0.1
Python: cpython-312 (3.12.7)
Platform: macOS-15.5-arm64-arm-64bit

### Python version
Python 3.12.7