description disappearing from Info when Uploading a Dataset Created with `from_dict`

### Describe the bug

When uploading a dataset created locally using `from_dict` with a specified `description` field. It appears before upload, but is missing after upload and re-download.

### Steps to reproduce the bug

I think the most relevant pattern in the code might be the following lines:

```
description_json_str = json.dumps(
{
"dataset_id": dataset.spec.dataset_id,
"env_name": dataset.spec.env_spec.id,
"action_space": serialize_space(dataset.spec.action_space),
"observation_space": serialize_space(dataset.spec.observation_space),
}
)

hugging_face_dataset = Dataset.from_dict(
episodes_dict, info=DatasetInfo(description=description_json_str)
)

```
Which comes from this function https://github.com/balisujohn/minarai/blob/8e023727f0a8488c4451651d9f7a79b981412c40/minari/integrations/hugging_face.py#L39

To replicate,
clone this branch of my Minari fork https://github.com/balisujohn/minarai/tree/dev-huggingface then run

```
python3.8 -m venv env
source env/bin/activate
python3 -m pip install -e .
python3 -m pip install pytest
```

The change the hugging face repo path in the test called `test_hugging_face_push_and_pull_dataset` in `tests/integrations/test_hugging_face.py` to one you have permissions to write to.

Then run:

```
pytest tests/integrations/test_hugging_face.py::test_hugging_face_push_and_pull_dataset
```

### Expected behavior

DATASET INFO BEFORE UPLOADING
DatasetInfo(description='{"dataset_id": "dummy-combo-test-v0", "env_name": "DummyComboEnv-v0", "action_space": "{\\"type\\": \\"Tuple\\", \\"subspaces\\": [{\\"type\\": \\"Box\\", \\"dtype\\": \\"float32\\", \\"shape\\": [1], \\"low\\": [2.0], \\"high\\": [3.0]}, {\\"type\\": \\"Box\\", \\"dtype\\": \\"float32\\", \\"shape\\": [1], \\"low\\": [4.0], \\"high\\": [5.0]}]}", "observation_space": "{\\"type\\": \\"Tuple\\", \\"subspaces\\": [{\\"type\\": \\"Box\\", \\"dtype\\": \\"float32\\", \\"shape\\": [1], \\"low\\": [2.0], \\"high\\": [3.0]}, {\\"type\\": \\"Tuple\\", \\"subspaces\\": [{\\"type\\": \\"Box\\", \\"dtype\\": \\"float32\\", \\"shape\\": [1], \\"low\\": [2.0], \\"high\\": [3.0]}, {\\"type\\": \\"Dict\\", \\"subspaces\\": {\\"component_1\\": {\\"type\\": \\"Box\\", \\"dtype\\": \\"float32\\", \\"shape\\": [1], \\"low\\": [-1.0], \\"high\\": [1.0]}, \\"component_2\\": {\\"type\\": \\"Dict\\", \\"subspaces\\": {\\"subcomponent_1\\": {\\"type\\": \\"Box\\", \\"dtype\\": \\"float32\\", \\"shape\\": [1], \\"low\\": [2.0], \\"high\\": [3.0]}, \\"subcomponent_2\\": {\\"type\\": \\"Tuple\\", \\"subspaces\\": [{\\"type\\": \\"Box\\", \\"dtype\\": \\"float32\\", \\"shape\\": [1], \\"low\\": [4.0], \\"high\\": [5.0]}, {\\"type\\": \\"Discrete\\", \\"dtype\\": \\"int64\\", \\"start\\": 0, \\"n\\": 10}]}}}}}]}]}"}', citation='', homepage='', license='', features={'observations': {'_index_0': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), '_index_1': {'_index_0': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), '_index_1': {'component_1': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), 'component_2': {'subcomponent_1': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), 'subcomponent_2': {'_index_0': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), '_index_1': Value(dtype='int64', id=None)}}}}}, 'actions': {'_index_0': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), '_index_1': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None)}, 'rewards': Value(dtype='int64', id=None), 'truncations': Value(dtype='bool', id=None), 'terminations': Value(dtype='bool', id=None), 'episode_ids': Value(dtype='int64', id=None)}, post_processed=None, supervised_keys=None, task_templates=None, builder_name=None, config_name=None, version=None, splits=None, download_checksums=None, download_size=None, post_processing_size=None, dataset_size=None, size_in_bytes=None)
...
DATASET INFO AFTER UPLOADING AND DOWNLOADING
DatasetInfo(description='', citation='', homepage='', license='', features={'observations': {'_index_0': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), '_index_1': {'_index_0': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), '_index_1': {'component_1': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), 'component_2': {'subcomponent_1': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), 'subcomponent_2': {'_index_0': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), '_index_1': Value(dtype='int64', id=None)}}}}}, 'actions': {'_index_0': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None), '_index_1': Sequence(feature=Value(dtype='float32', id=None), length=-1, id=None)}, 'rewards': Value(dtype='int64', id=None), 'truncations': Value(dtype='bool', id=None), 'terminations': Value(dtype='bool', id=None), 'episode_ids': Value(dtype='int64', id=None)}, post_processed=None, supervised_keys=None, task_templates=None, builder_name=None, config_name=None, version=None, splits={'train': SplitInfo(name='train', num_bytes=4846, num_examples=60, shard_lengths=None, dataset_name='parquet')}, download_checksums={'https://huggingface.co/datasets/balisujohn/minari_test/resolve/8217b614ff9ba5edc1a30c7df430e92a46f65363/data/train-00000-of-00001-7c5900b93b35745e.parquet': {'num_bytes': 9052, 'checksum': None}}, download_size=9052, post_processing_size=None, dataset_size=4846, size_in_bytes=13898)
...

### Environment info

- `datasets` version: 2.13.0
- Platform: Linux-5.15.0-75-generic-x86_64-with-glibc2.29
- Python version: 3.8.10
- Huggingface_hub version: 0.15.1
- PyArrow version: 12.0.1
- Pandas version: 2.0.2