Strange slicing behavior

### Describe the bug

I have loaded a dataset, and then slice first 300 samples using `:` ops, however, the resulting dataset is not expected, as the output below:

```bash
len(dataset)=1050324
len(dataset[:300])=2
len(dataset[0:300])=2
len(dataset.select(range(300)))=300
```

### Steps to reproduce the bug

load a dataset then:
```bash
dataset = load_from_disk(args.train_data_dir)

print(f"{len(dataset)=}", flush=True)
print(f"{len(dataset[:300])=}", flush=True)
print(f"{len(dataset[0:300])=}", flush=True)
print(f"{len(dataset.select(range(300)))=}", flush=True)
```

### Expected behavior

```bash
len(dataset)=1050324
len(dataset[:300])=300
len(dataset[0:300])=300
len(dataset.select(range(300)))=300
```

### Environment info

- `datasets` version: 2.16.1
- Platform: Linux-5.15.0-60-generic-x86_64-with-glibc2.35
- Python version: 3.10.11
- `huggingface_hub` version: 0.20.2
- PyArrow version: 10.0.1
- Pandas version: 1.5.3
- `fsspec` version: 2023.10.0