Bug loading `sentence-transformers/parallel-sentences`

## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("sentence-transformers/parallel-sentences")
```

raises this:

```
/home/phmay/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/download/streaming_download_manager.py:697: FutureWarning: the 'mangle_dupe_cols' keyword is deprecated and will be removed in a future version. Please take steps to stop the use of 'mangle_dupe_cols'
return pd.read_csv(xopen(filepath_or_buffer, "rb", use_auth_token=use_auth_token), **kwargs)
/home/phmay/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/download/streaming_download_manager.py:697: FutureWarning: the 'mangle_dupe_cols' keyword is deprecated and will be removed in a future version. Please take steps to stop the use of 'mangle_dupe_cols'
return pd.read_csv(xopen(filepath_or_buffer, "rb", use_auth_token=use_auth_token), **kwargs)

---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In [4], line 1
----> 1 dataset = load_dataset("sentence-transformers/parallel-sentences", split="train")

File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/load.py:1693, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, **config_kwargs)
1690 try_from_hf_gcs = path not in _PACKAGED_DATASETS_MODULES
1692 # Download and prepare data
-> 1693 builder_instance.download_and_prepare(
1694 download_config=download_config,
1695 download_mode=download_mode,
1696 ignore_verifications=ignore_verifications,
1697 try_from_hf_gcs=try_from_hf_gcs,
1698 use_auth_token=use_auth_token,
1699 )
1701 # Build dataset for splits
1702 keep_in_memory = (
1703 keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)
1704 )

File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/builder.py:807, in DatasetBuilder.download_and_prepare(self, output_dir, download_config, download_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, file_format, max_shard_size, storage_options, **download_and_prepare_kwargs)
801 if not downloaded_from_gcs:
802 prepare_split_kwargs = {
803 "file_format": file_format,
804 "max_shard_size": max_shard_size,
805 **download_and_prepare_kwargs,
806 }
--> 807 self._download_and_prepare(
808 dl_manager=dl_manager,
809 verify_infos=verify_infos,
810 **prepare_split_kwargs,
811 **download_and_prepare_kwargs,
812 )
813 # Sync info
814 self.info.dataset_size = sum(split.num_bytes for split in self.info.splits.values())

File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/builder.py:898, in DatasetBuilder._download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)
894 split_dict.add(split_generator.split_info)
896 try:
897 # Prepare split will record examples associated to the split
--> 898 self._prepare_split(split_generator, **prepare_split_kwargs)
899 except OSError as e:
900 raise OSError(
901 "Cannot find data file. "
902 + (self.manual_download_instructions or "")
903 + "\nOriginal error:\n"
904 + str(e)
905 ) from None

File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/builder.py:1513, in ArrowBasedBuilder._prepare_split(self, split_generator, file_format, max_shard_size)
1506 shard_id += 1
1507 writer = writer_class(
1508 features=writer._features,
1509 path=fpath.replace("SSSSS", f"{shard_id:05d}"),
1510 storage_options=self._fs.storage_options,
1511 embed_local_files=embed_local_files,
1512 )
-> 1513 writer.write_table(table)
1514 finally:
1515 num_shards = shard_id + 1

File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/arrow_writer.py:540, in ArrowWriter.write_table(self, pa_table, writer_batch_size)
538 if self.pa_writer is None:
539 self._build_writer(inferred_schema=pa_table.schema)
--> 540 pa_table = table_cast(pa_table, self._schema)
541 if self.embed_local_files:
542 pa_table = embed_table_storage(pa_table)

File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/table.py:2044, in table_cast(table, schema)
2032 """Improved version of pa.Table.cast.
2033
2034 It supports casting to feature types stored in the schema metadata.
(...)
2041 table (:obj:`pyarrow.Table`): the casted table
2042 """
2043 if table.schema != schema:
-> 2044 return cast_table_to_schema(table, schema)
2045 elif table.schema.metadata != schema.metadata:
2046 return table.replace_schema_metadata(schema.metadata)

File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/table.py:2005, in cast_table_to_schema(table, schema)
2003 features = Features.from_arrow_schema(schema)
2004 if sorted(table.column_names) != sorted(features):
-> 2005 raise ValueError(f"Couldn't cast\n{table.schema}\nto\n{features}\nbecause column names don't match")
2006 arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
2007 return pa.Table.from_arrays(arrays, schema=schema)

ValueError: Couldn't cast
Action taken on Parliament's resolutions: see Minutes: string
Následný postup na základě usnesení Parlamentu: viz zápis: string
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 742
to
{'Membership of Parliament: see Minutes': Value(dtype='string', id=None), 'Състав на Парламента: вж. протоколи': Value(dtype='string', id=None)}
because column names don't match
```

## Expected results
no error

## Actual results
error

## Environment info

- `datasets` version:
- Platform: Linux
- Python version: Python 3.9.13
- PyArrow version: pyarrow 9.0.0
- transformers 4.22.2
- datasets 2.5.2