Provide a temporary implementation for `array_api.compat.torch.unique_all` instead of raising `NotImplementedError`

[ogrisel]

The first snippet provided in this comment pytorch/pytorch#36748 (comment) might help smooth support for torch in Array API consuming libraries such as scikit-learn without them having to maintain their own hacks while waiting upstream PyTorch to address pytorch/pytorch#70920.

[asmeurer]

This sounds reasonable. Based on the code comments, the workarounds there didn't work because they used non-deterministic scatter. But if those workarounds you referenced don't have that issue, we can implement them.