texts extracted from different rows of table

### Bug

...
'2024 年 12 月' recognized as '2024 年 12 月 31 日 1 5 年以上'.

[成都银行2024年度年报252.pdf](https://github.com/user-attachments/files/20682032/2024.252.pdf)

### Steps to reproduce

...DOC_SOURCE = "/media/will/D/RAG-Challenge-2/pdf_reports/成都银行2024年度年报252.pdf"

# we set some start-stop cues for defining an excerpt to print
start_cue = "本集团"
stop_cue = "页"
from rich.console import Console
from rich.panel import Panel

console = Console(width=210) # for preventing Markdown table wrapped rendering

def print_in_console(text):
console.print(Panel(text))
from docling.document_converter import DocumentConverter

converter = DocumentConverter()
doc = converter.convert(source=DOC_SOURCE).document

### Docling version

...

### Python version

...docling 2.36.1
docling-core 2.34.1
docling-ibm-models 3.4.4
docling-parse 4.0.3

Start with the supplied 成都银行2024年度年报252.pdf reproducer and the DocumentConverter call, then inspect the extracted text between the “本集团” and “页” cues. Confirm how the table rows become interleaved and compare the output around “2024 年 12 月”. Done means text from separate rows is no longer combined in the extracted result.

Written by the indexing model from the issue text.