Introduce torch.compile to all PyTorch examples

[champon1020]

As the previous discussion #2024 (comment), I would suggest to introduce torch.compile in PyTorch examples, following the migration of PyTorch to v2 (#2024).

I think it should be guaranteed to work torch.compile, which is the main feature of PyTorch v2 and able to speed up computation and memory access, on PyTorch operator.
However, as @tenzen-y mentioned at #2024 (comment), there is a concern about supported GPUs (or I think also CUDA versions).

Please let me know your opinions!