load_datasets("json", ...) don't read local .json.gz properly

## Describe the bug
I have a local file `*.json.gz` and it can be read by `pandas.read_json(lines=True)`, but cannot be read by `load_datasets("json")` (resulting in 0 lines)

## Steps to reproduce the bug
```python
fpath = '/data/junwang/.cache/general/57b6f2314cbe0bc45dda5b78f0871df2/test.json.gz'
ds_panda = DatasetDict(
test=Dataset.from_pandas(
pd.read_json(fpath, lines=True)
)
)
ds_direct = load_dataset(
'json', data_files={
'test': fpath
}, features=Features(
text_input=Value(dtype="string", id=None),
text_output=Value(dtype="string", id=None)
)
)
len(ds_panda['test']), len(ds_direct['test'])
```

## Expected results
Lines of `ds_panda['test']` and `ds_direct['test']` should match.

## Actual results
```
Using custom data configuration default-c0ef2598760968aa
Downloading and preparing dataset json/default to /data/junwang/.cache/huggingface/datasets/json/default-c0ef2598760968aa/0.0.0/e6070c77f18f01a5ad4551a8b7edfba20b8438b7cad4d94e6ad9378022ce4aab...
Dataset json downloaded and prepared to /data/junwang/.cache/huggingface/datasets/json/default-c0ef2598760968aa/0.0.0/e6070c77f18f01a5ad4551a8b7edfba20b8438b7cad4d94e6ad9378022ce4aab. Subsequent calls will reuse this data.

(62087, 0)
```

## Environment info

- `datasets` version:
- Platform: Ubuntu 18.04.4 LTS
- Python version: 3.8.13
- PyArrow version: 9.0.0