TableFormer fails on borderless tables with whitespace-delimited columns

### Bug
Docling doesn't detect tables that have no visible borders — only whitespace-aligned columns. The text extracts fine (all words present, readable), but the table structure is completely missed; content gets treated as body text instead of rows/columns.

Common in financial documents and technical PDFs where borders are intentionally omitted and column alignment is the only structural signal.
...

### Steps to reproduce
- PDF with a borderless table (columns aligned by whitespace, no visible lines)
- Run Docling extraction
- Table region not identified in output
...

### Docling version
latest from pip
...

### Python version
3.11
...

Start by reproducing the report with a PDF containing whitespace-aligned, borderless columns and inspect the extraction output to confirm that text is present but no table structure is emitted. The issue names no source file or test entry point, so trace the table-detection path used by Docling and identify where borderless layouts are handled. Done means such tables are recognized as rows and columns without regressing bordered-table extraction.

Written by the indexing model from the issue text.