Fix tensor parallel distributed tests by requesting PyTorch not to destroy PG upon exit

[IvanYashchuk]

Before this PR tensor parallel tests were failing. Example

pytest thunder/tests/distributed/test_tensor_parallel.py::TensorParallelTest::test_embedding_name_column

ValueError: Default process group has not been initialized, please make sure to call init_process_group.
Process process 1:
Traceback (most recent call last):
  File "/usr/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/usr/lib/python3.12/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/opt/pytorch/lightning-thunder/thunder/tests/distributed/helper.py", line 144, in _run
    torch.distributed.barrier()
  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/c10d_logger.py", line 81, in wrapper
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/distributed_c10d.py", line 4462, in barrier
    opts.device = torch.device(_get_object_coll_device(group))
                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/distributed_c10d.py", line 774, in _get_object_coll_device
    group = group or _get_default_group()
                     ^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/distributed/distributed_c10d.py", line 1276, in _get_default_group
    raise ValueError(
ValueError: Default process group has not been initialized, please make sure to call init_process_group.

And now it works.

cc @Borda @crcrpar

[t-vi]

Thank you Ivan, this is fixed in #1470 already along with the other.