fix(BEE): recover unmatched tokens in table last row when model under-predicts row count

[AnkitAhlawat7742]

Fix #3402

issue description
When the last row of a table sits at the very bottom edge of the table image crop , the TableFormer model fails to
predict it as a separate row. The tokens are present in the input token list and well within the page bounds, but the model simply doesn't output a row for them.

Resolves

This solution adds a post-processing recovery step in TableStructureModel.predict_tablesthat detects word-level tokens not assigned to any predicted cell and appends them as an extra trailing row.

[github-actions]

✅ DCO Check Passed

Thanks @AnkitAhlawat7742, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

🟢 Merge protection satisfied — ready to merge.

Show 1 satisfied protection

🟢 Enforce conventional commit

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)(?:\(.+\))?(!)?:

[cau-git]

@AnkitAhlawat7742 Thanks for the proposal. Before we resort to recovery-style post-processing, could you please check instead if enlarging the table detection a bit towards the bottom would cover this as well? Then we can see if that has undesired side effects on previously working tables and decide.