Native PDF content parsed as images (output contains only <!-- image -->)

### Bug

I am attempting to parse a PDF document using Docling. The PDF in question is a native digital document (not a scan), and the text is selectable using standard PDF viewers.

However, when running the conversion, Docling fails to extract the text content. Instead, the Markdown output consists solely of `` placeholders, effectively treating the pages or content blocks as unparsed images.

I expected Docling to extract the text and structure, given that the file is not a rasterized scan.

### Steps to reproduce

Use the attached PDF: [Shifts Test Docling Issue.pdf](https://github.com/user-attachments/files/25340696/Shifts.Test.Docling.Issue.pdf)

Setup: docling[rapidocr] with ACCURATE or FAST mode and these options:

```
def get_pipeline_options() -> PdfPipelineOptions:
"""
Creates and returns the shared PdfPipelineOptions.
Default mode is ACCURATE.
"""
pipeline_options = PdfPipelineOptions()
pipeline_options.do_ocr = True
pipeline_options.images_scale = 4.0 # High res for small text inside images

pipeline_options.do_table_structure = True
pipeline_options.table_structure_options.mode = TableFormerMode.ACCURATE
pipeline_options.ocr_options = RapidOcrOptions()
pipeline_options.generate_picture_images = True
pipeline_options.table_structure_options.do_cell_matching = True

return pipeline_options
```

### Docling version

2.73.1

### Python version

3.13