`Dataset` unexpected changed dict data and may cause error

### Describe the bug

Will unexpected get keys with `None` value in the parsed json dict.

### Steps to reproduce the bug

```jsonl test.jsonl
{"id": 0, "indexs": {"-1": [0, 10]}}
{"id": 1, "indexs": {"-1": [0, 10]}}
```

```python
dataset = Dataset.from_json('.test.jsonl')
print(dataset[0])
```

Result:
```
{'id': 0, 'indexs': {'-1': [...], '-2': None, '-3': None, '-4': None, '-5': None, '-6': None, '-7': None, '-8': None, '-9': None, ...}}
```

Those keys with `None` value will unexpected appear in the dict.

### Expected behavior

Result should be
```
{'id': 0, 'indexs': {'-1': [0, 10]}}
```

### Environment info

- `datasets` version: 2.16.1
- Platform: Linux-6.5.0-14-generic-x86_64-with-glibc2.35
- Python version: 3.11.6
- `huggingface_hub` version: 0.20.2
- PyArrow version: 14.0.2
- Pandas version: 2.1.4
- `fsspec` version: 2023.10.0