If an image in a dataset is corrupted, we get unescapable error

### Describe the bug

Example discussed in detail here: https://huggingface.co/datasets/sasha/birdsnap/discussions/1

### Steps to reproduce the bug

```
from datasets import load_dataset, VerificationMode
dataset = load_dataset(
'sasha/birdsnap',
split="train",
verification_mode=VerificationMode.ALL_CHECKS,
streaming=True # I recommend using streaming=True when reproducing, as this dataset is large
)
for idx, row in enumerate(dataset):
# Iterating to 9287 took 7 minutes for me
# If you already have the data locally cached and set streaming=False, you see the same error just by with dataset[9287]
pass
# error at 9287 OSError: image file is truncated (45 bytes not processed)
# note that we can't avoid the error using a try/except + continue inside the loop
```

### Expected behavior

Able to escape errors in casting to Image() without killing the whole loop

### Environment info

- `datasets` version: 2.15.0
- Platform: Linux-5.15.0-84-generic-x86_64-with-glibc2.31
- Python version: 3.11.5
- `huggingface_hub` version: 0.19.4
- PyArrow version: 14.0.1
- Pandas version: 2.1.3
- `fsspec` version: 2023.10.0