[Bee] Docling disable layout model or use fast custom layout model to support express mode

### Requested feature
IBM WDU project plans to use Docling for text extraction in express mode, the expectation is to process 1000 pages of programmatic PDF under 60 seconds. However, current docling takes about 120 seconds because the layout mode processing cannot be disabled.
- backend: DoclingParseV4DocumentBackend (default). If I switch to PyPdfiumDocumentBackend, only a few seconds less than the default.
- output: only plain text is required

The summary of overall timing:
> Extracted timing for docling_pipeline_total: 114.299s (count=1)
> Extracted timing for docling_page_parse: 25.766s (count=1000)
> Extracted timing for docling_layout: 98.511s (count=251)
> Extracted timing for docling_page_assemble: 0.079s (count=1000)
> Extracted timing for docling_doc_assemble: 5.286s (count=1)
> Extracted timing for docling_reading_order: 5.264s (count=1)
> Extracted timing for docling_doc_enrich: 1.162s (count=1)

test doc:
[test_pdf_1000.pdf](https://github.ibm.com/user-attachments/files/1733021/test_pdf_1000.pdf)