Make index_streaming_dataset take in an IterableDataset

[cakiki]

Current signature is: dataset_name_or_path: str. I think it would be better to separate streaming a dataset from streaming the indexing and have the function take a datasets.IterableDataset object as input because that's what's returned from the data subpackage. e.g. when loading a streaming dataset from ir_datasets with: hf_dset_streaming = load_ir_dataset_streaming(IR_DSET).

WDYT @ToluClassics @theyorubayesian

[theyorubayesian]

Doesn't sound too difficult to implement & it's a separation of concerns so it makes sense