Fix: preserve features after IterableDataset.add_column (GH#5752)

[uttam12331]

Problem

After calling .add_column() on a streaming IterableDataset, accessing .features returns None:

from datasets import load_dataset

ds = load_dataset("...", streaming=True)
print(ds.features)  # {'text': Value('string'), ...}

ds2 = ds.add_column("new_column", ["val"] * len(ds))
print(ds2.features)  # None  <-- bug
print(ds2.features.keys())  # AttributeError: 'NoneType' object has no attribute 'keys'

Root Cause

add_column calls self.map(...) without passing the features argument. Inside _map, the new dataset's info is constructed as:

info = self.info.copy()
info.features = features  # None if not passed

This unconditionally overwrites the copied features with None.

Fix

Infer the feature type of the new column from the provided data, merge it with the existing features, and pass the combined Features object to map():

column_features = _infer_features_from_batch({name: list(column) if not isinstance(column, list) else column})
if self.features is not None:
    features = self.features.copy()
    features.update(column_features)
else:
    features = column_features
return self.map(..., features=features)

Tests

• Added test_iterable_dataset_add_column_preserves_features which explicitly checks that .features is not None and contains both original and new columns after add_column.
• Existing test_iterable_dataset_add_column continues to pass.

Closes #5752