`wmt_t2t_translate` fails build when `--data_dir` set to GCS location

/!\ PLEASE INCLUDE THE FULL STACKTRACE AND CODE SNIPPET

Short description
wmt_t2t_translate will not build when --data_dir is set to a GCS location due to etils.Path().glob() behavior when glob pattern is a gcs URI

Environment information

• Operating System: Linux

• Python version: 3.9.15

• tensorflow-datasets/tfds-nightly version: 4.8.2 / 4.8.2.dev202301270045

• tensorflow/tf-nightly version: 2.11.0

• Does the issue still exists with the last tfds-nightly package (pip install --upgrade tfds-nightly) ?

• yes

Reproduction instructions

• pip install tensorflow tensorflow-datasets
• tfds build --data_dir $GCS_DATA_DIR wmt_t2t_translate

INFO[wmt.py]: Using sub-datasets: {Split('train'): ['europarl_v7', 'commoncrawl', 'newscommentary_v13'], Split('validation'): ['newstest2013'], Split('test'): ['newstest2014']}
INFO[wmt.py]: Generating examples from: europarl_v7                                                                                                                                                                                                                                                                                                                                                                                   
Traceback (most recent call last):                                                                                                                                                                                                                                                                                                                                                                                                    
  File "/opt/conda/envs/tfds-bug/bin/tfds", line 8, in <module>                                                                                                                                                                                                                                                                                                                                                                       
    sys.exit(launch_cli())
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/scripts/cli/main.py", line 109, in launch_cli
    app.run(main, flags_parser=_parse_flags)
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/absl/app.py", line 308, in run
    _run_main(main, args)
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/absl/app.py", line 254, in _run_main
    sys.exit(main(argv))
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/scripts/cli/main.py", line 104, in main
    args.subparser_fn(args)
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/scripts/cli/build.py", line 274, in _build_datasets
    _download_and_prepare(args, builder)
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/scripts/cli/build.py", line 510, in _download_and_prepare
    builder.download_and_prepare(
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/core/logging/__init__.py", line 169, in __call__
    return function(*args, **kwargs)
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/core/dataset_builder.py", line 642, in download_and_prepare
    self._download_and_prepare(
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/core/dataset_builder.py", line 1487, in _download_and_prepare
    future = split_builder.submit_split_generation(
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/core/split_builder.py", line 340, in submit_split_generation
    return self._build_from_generator(**build_kwargs)
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/core/split_builder.py", line 406, in _build_from_generator
    for key, example in utils.tqdm(
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tqdm/std.py", line 1195, in __iter__
    for obj in iterable:
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/translate/wmt.py", line 938, in _generate_examples
    for sub_key, ex in sub_generator(*files):
  File "/opt/conda/envs/tfds-bug/lib/python3.9/site-packages/tensorflow_datasets/translate/wmt.py", line 1000, in _parse_parallel_sentences
    assert f1_files and f2_files, "No matching files found: %s, %s." % (f1, f2)
AssertionError: No matching files found: gs://gcp-ml-sandbox-t5x/datasets/downloads/extracted/TAR_GZ.statmt.org_wmt13_traini-parall-europa-v7AiTHxxDIoGPf2JOwzAgwIC1h9MdcF-uOMYNhA9J9luc.tgz/training/europarl-v7.de-en.de, gs://gcp-ml-sandbox-t5x/datasets/downloads/extracted/TAR_GZ.statmt.org_wmt13_traini-parall-europa-v7AiTHxxDIoGPf2JOwzAgwIC1h9MdcF-uOMYNhA9J9luc.tgz/training/europarl-v7.de-en.en.
(tfds-bug) jupyter@python3-conda-gpu:~/T5X$ 

If you share a colab, make sure to update the permissions to share it.

Link to logs
If applicable, <link to gist with logs, stack trace>

Expected behavior
dataset to be built

Additional context
Seems to be an issue with file globbing behavior being different for gcs vs local paths

Start with the tfds build --data_dir $GCS_DATA_DIR wmt_t2t_translate reproduction and read tensorflow_datasets/translate/wmt.py, especially _parse_parallel_sentences near the failing assertion. Compare how etils.Path().glob() handles the GCS paths shown in the traceback versus local paths. Done means the dataset builds successfully with a GCS --data_dir without the no-matching-files assertion.