feat: update OmniDocBench with Parquet support and add test

[samiuc]

The OmniDocBench dataset contains ~2700 individual files. When downloading via the raw mode (file-by-file), each file triggers a separate request to HuggingFace's API. In parallel test environments, this quickly hits HuggingFace's rate limit of 1000 requests per 5 minutes, causing test failures.

So in the PR, I have:

• Added a new use_parquet parameter to OmniDocBenchDatasetBuilder that enables downloading pre-converted Parquet shards instead of individual files
• Implemented _iterate_parquet() method that uses load_dataset() to stream data from Parquet files, reducing the number of API requests
• Parquet mode downloads the entire dataset instead of 2700+ individual requests
• Maintains backward compatibility with the original raw mode (default behavior unchanged)

The Parquet version of OmniDocBench is available at: https://huggingface.co/datasets/samiuc/OmniDocBench-parquet

Question: can we move it to docling repo on HF?

[github-actions]

✅ DCO Check Passed

Thanks @samiuc, 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

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

[divekarsc]

LGTM!

[samiuc]

@cau-git can you please review the PR when you have sometime? Thank you