Bad split error message when using percentages

### Describe the bug

Hi, I'm trying to download a dataset. To not load the entire dataset in memory, I split it as described [here](https://huggingface.co/docs/datasets/v4.0.0/loading#slice-splits) in 10% steps.

When doing so, the library returns this error:
raise ValueError(f"Bad split: {split}. Available splits: {list(splits_generators)}")
ValueError: Bad split: train[0%:10%]. Available splits: ['train']

Edit: Same happens with a split like _train[:90000]_
### Steps to reproduce the bug

```
for split in range(10):
split_str = f"train[{split*10}%:{(split+1)*10}%]"
print(f"Processing split {split_str}...")
ds = load_dataset("user/dataset", split=split_str, streaming=True)
```

### Expected behavior

I'd expect the library to split my dataset in 10% steps.

### Environment info

python 3.12.11
ubuntu 24
dataset 4.0.0