fix(serializer): stop duplicating spanning cells in markdown/dataframe output

[scottf007]

Summary

The markdown table serializer and the legacy markdown / dataframe export paths iterate data.grid and emit each grid cell's text. When a TableCell has col_span > 1 or row_span > 1, the same cell reference occupies every grid position it covers, so iterating produces N copies of the cell's text in the output instead of one.

The underlying TableCell data is correct (TableFormer emits one cell per merged region with col_span / row_span set). Only the serialized presentation duplicates.

Visible symptoms

• A col_span=4 header like `KRYTERIA FORMALNE` rendered as four repeated cells instead of one cell + three empties.
• A row_span=2 cell like `Energy reduction` rendered twice in consecutive rows of the markdown table.
• export_to_dataframe() concatenated the parent header text with each child header (e.g. `KRYTERIA FORMALNE.Lp.`, `KRYTERIA FORMALNE.TAK`).

Fix

In all three code paths:

• MarkdownTableSerializer.serialize (modern path)
• TableItem.export_to_markdown (legacy path)
• TableItem.export_to_dataframe

emit cell text only at the cell's origin position (start_row_offset_idx, start_col_offset_idx). Spanned positions render as empty string. Cells with col_span=1 and row_span=1 are unaffected because their single grid position equals their origin.

Test plan

• `pytest test/test_docling_doc.py` — 67 pass after fixture regen
• Full non-chunker test suite — 376 pass, 2 xfailed, no regressions
• Manual run via `docling` on three PDFs that exercise spans:
  • CVS Group 2025 annual report (TCFD table, page 35) — `Energy reduction` rowspan now renders once
  • Polish gov "Czyste Powietrze" — `KRYTERIA FORMALNE` colspan=4 now renders once
  • iShares dividend statement — `Net Cash for Reinvestment#` colspan=2 now renders once
• Regenerated 6 affected groundtruth fixtures (`.yaml.md`, `.paged.md`, `constructed_doc*.md.gt`). Diffs are precisely the removed cell-content duplication, nothing else.

Issues addressed

Closes docling-project/docling#2862 — `export_to_dataframe` produces duplicated cells when tables contain spanning cells. The dataframe path is fixed by this change.

Notes

• `data.grid` semantics are unchanged — the same `TableCell` reference still occupies multiple grid positions for spans. Only the serializers' use of it changes.
• HTML serializer wasn't modified; HTML supports native `colspan` / `rowspan` so the duplication issue may not apply there. Please flag if it does.

[github-actions]

✅ DCO Check Passed

Thanks @scottf007, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

Your pull request matches the following merge protections and will not be merged until they are valid.

🔴 Require two reviewer for test updates

Waiting for:

• #approved-reviews-by >= 2

This rule is failing.

When test data is updated, we require two reviewers

• #approved-reviews-by >= 2

🟢 Enforce conventional commit

Wonderful, this rule succeeded.

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?: