feat: Add h5folder dataset loader for HDF5 support

[ArjunJagdale]

Related Issue

Closes #3113

What does this PR do?

This PR introduces a new dataset loader module called h5folder to support loading datasets stored in HDF5 (.h5) format.

It allows users to do:

from datasets import load_dataset
dataset = load_dataset("h5folder", data_dir="path/to/")

🧩 Design Overview

• Implemented inside datasets/packaged_modules/h5folder/h5folder.py
• Based on the GeneratorBasedBuilder API
• Uses h5py to read HDF5 files and yield examples
• Expects datasets such as id, data, and label inside data.h5
• Converts numpy arrays to Python types before yielding

🧪 Example .h5 Structure (for local testing)

import h5py
import numpy as np

with h5py.File("data.h5", "w") as f:
    f.create_dataset("id", data=np.arange(100))
    f.create_dataset("data", data=np.random.randn(100, 10))
    f.create_dataset("label", data=np.random.randint(0, 2, size=100))

✅ Testing

• The loader logic follows the structure of existing modules like imagefolder
• Will rely on Hugging Face CI to validate integration
• Manually testing planned once merged or during feedback

📁 Files Added

• datasets/src/datasets/packaged_modules/h5folder/h5folder.py

📌 Component(s) Affected

• area/datasets
• area/load

📦 Release Note Classification

• rn/feature – Adds support for loading .h5 datasets via load_dataset("h5folder", ...)

---

Let me know if any changes or improvements are needed — happy to iterate. Thanks for reviewing!

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[ArjunJagdale]

I guess test failed cause import os, import h5py, and import datasets lines are not alphabetically sorted, or not grouped properly.

[ArjunJagdale]

This commit was accidental - [Merge branch 'main' into patch-4]. The
[chore: fix import order in h5folder.py to satisfy linter] should solve the import order issue.