fix: use weights_only=True in torch.load to prevent arbitrary code execution

[theeldruin]

Summary

`examples/sampling_pytorch.py:70` calls `torch.load(path_checkpoint)` without `weights_only=True`. PyTorch's default behavior uses pickle for deserialization, meaning a crafted `.pt` file can execute arbitrary code when loaded.

Change

```python

Before

params = torch.load(path_checkpoint)

After

params = torch.load(path_checkpoint, weights_only=True)
```

`weights_only=True` restricts deserialization to tensor types only, blocking `reduce` payloads entirely. It has been available since PyTorch 1.13 and is the recommended approach in PyTorch's own security guidance.

This was reported to Google via Bug Hunter and confirmed as a valid vulnerability.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.