Multi-column layout extraction fails - text flows incorrectly across columns despite table structure recognition enabled

### Bug
I'm processing financial documents with 3-column layouts using Docling, but the text extraction is flowing incorrectly across column boundaries, making the output unusable for RAG applications.
...

### Steps to reproduce
Expected Behavior:
Text should be extracted respecting column boundaries and reading order, with content from each column kept separate and in proper sequence.
Actual Behavior:
Text flows left-to-right across all columns ignoring boundaries, creating scrambled content where sentences from different columns are merged together.

Is there a specific configuration needed for proper multi-column layout recognition? The documentation suggests this should work with table structure recognition enabled, but column boundaries are not being respected.

```
from docling.datamodel.pipeline_options import PdfPipelineOptions, TableFormerMode

pdf_pipeline_options = PdfPipelineOptions(
do_ocr=False,
generate_picture_images=True,
do_table_structure=True, # Enabled for layout analysis
do_picture_description=True
)

pdf_pipeline_options.table_structure_options.do_cell_matching = False
pdf_pipeline_options.table_structure_options.mode = TableFormerMode.ACCURATE
```
...

### Docling version
Docling version: 2.43.0
Docling Core version: 2.44.1
Docling IBM Models version: 3.9.0
Docling Parse version: 4.1.0
Platform: Linux-6.8.0-1031-azure-x86_64-with-glibc2.35
...

### Python version
Python: cpython-310 (3.10.12)
...