fix: preserve features in IterableDataset.add_column (fixes #5752)

[IgnazioDS]

Summary

Fixes #5752.

IterableDataset.add_column() lost the dataset's feature schema on every call because it delegated to map() without passing the features argument. map() unconditionally writes info.features = features, so the default None silently overwrote the existing schema.

Root cause (iterable_dataset.py:3459):

info = self.info.copy()
info.features = features  # None when add_column() doesn't pass features

Fix: infer the new column's Arrow type via the existing _infer_features_from_batch helper, merge it with the existing features, and pass the result to map().

new_features = None
if self._info.features is not None:
    column_features = _infer_features_from_batch({name: list(column)})
    new_features = Features({**self._info.features, **column_features})
return self.map(..., features=new_features)

If the dataset had no features to begin with (_info.features is None), behaviour is unchanged — map() receives None as before.

Test plan

• Reproduce the original issue: create a streaming dataset with known features, call .add_column(), assert .features is not None and contains the new column
• Verify existing test_iterable_dataset.py tests still pass
• Smoke-test with numpy array column input (not just list)

[lhoestq]

lgtm ! I just added a suggestion, lmk what you think before we merge

[lhoestq]

this could help avoid unnecessary data copies, since _infer_features_from_batch does copy all the data into arrow

Suggested change

	column_features = _infer_features_from_batch({name: list(column)})
	column_features = _infer_features_from_batch({name: list(column[:config.DEFAULT_MAX_BATCH_SIZE])})

[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.