All layout models produce empty markdown output with no text content

### Bug
All Docling layout models (EGRET_MEDIUM, EGRET_LARGE, EGRET_XLARGE, V2, HERON_101) produce malformed markdown output with empty headers, bullets, and image placeholders when converting PDFs using `export_to_markdown()`.

The output consists primarily of:
- Empty markdown headers (`## \n\n`)
- Empty bullet points (`- \n`)
- Image placeholders (`\n\n`)
- Empty table structures
- No actual text content extracted from the PDF

### Steps to reproduce

1. Set up DocumentConverter with any layout model:

```python
from docling.document_converter import DocumentConverter, PdfFormatOption
from docling.datamodel.pipeline_options import PdfPipelineOptions, LayoutOptions, AcceleratorOptions
from docling.datamodel.base_models import InputFormat
from docling.datamodel.accelerator_options import AcceleratorDevice
from docling.datamodel.layout_model_specs import (
DOCLING_LAYOUT_EGRET_MEDIUM,
DOCLING_LAYOUT_EGRET_LARGE,
DOCLING_LAYOUT_EGRET_XLARGE,
DOCLING_LAYOUT_V2,
DOCLING_LAYOUT_HERON,
DOCLING_LAYOUT_HERON_101
)

# Configure pipeline options
_pipeline_options = PdfPipelineOptions()
_pipeline_options.generate_picture_images = True
_pipeline_options.images_scale = 1.0
_pipeline_options.do_ocr = False
_pipeline_options.do_table_structure = False
_pipeline_options.do_picture_classification = False
_pipeline_options.do_picture_description = False

_pipeline_options.accelerator_options = AcceleratorOptions(
device=AcceleratorDevice.CPU,
num_threads=8,
)

_pipeline_options.layout_options = LayoutOptions(
model_spec=DOCLING_LAYOUT_EGRET_LARGE, # Tried all models - same issue
skip_cell_assignment=True,
create_orphan_clusters=False,
)

# Create converter
doc_converter = DocumentConverter(
format_options={
InputFormat.PDF: PdfFormatOption(pipeline_options=_pipeline_options)
}
)
```

2. Convert a PDF document:

```python
from docling.datamodel.base_models import DocumentStream
from io import BytesIO

# Read PDF file
with open("sample.pdf", "rb") as f:
pdf_bytes = f.read()

buf = BytesIO(pdf_bytes)
source = DocumentStream(name="sample.pdf", stream=buf)

# Convert
result = doc_converter.convert(source)
doc = result.document

# Export to markdown
content = doc.export_to_markdown()
print(content)
```

3. Observe malformed output:

```markdown

##

##

##

##

##

##

##

##

##

##

##

##

-
-

##

##

-
-
-

##

##

##

| ``` ``` |
|------------|

##

##

```

### Expected Behavior
The markdown output should contain:
- Actual text content from the PDF
- Properly formatted headers with text
- Bullet points with content
- Table data (if present)
- Figure captions and descriptions

### Actual Behavior
The markdown output contains only empty structural elements with no text content.

### Additional Context
- **Tested Models**: All available layout models produce the same malformed output expect for the default model:
- `DOCLING_LAYOUT_EGRET_MEDIUM`
- `DOCLING_LAYOUT_EGRET_LARGE`
- `DOCLING_LAYOUT_EGRET_XLARGE`
- `DOCLING_LAYOUT_V2`
- `DOCLING_LAYOUT_HERON_101`

- **Pipeline Settings Tested**:
- OCR disabled: `do_ocr = False`
- Table structure disabled: `do_table_structure = False`
- Layout options: `skip_cell_assignment=True`, `create_orphan_clusters=False`

- **Environment**:
- OS: Windows
- Running in FastAPI service with async execution
- Using ThreadPoolExecutor for document conversion

### Docling version
docling 2.65.0

### Python version
Python 3.12.4