Distributed launch

pylibwholegraph.torch.add_distributed_launch_options(parser)
pylibwholegraph.torch.distributed_launch(...)
pylibwholegraph.torch.get_rank()
pylibwholegraph.torch.get_world_size()
pylibwholegraph.torch.get_local_rank()
pylibwholegraph.torch.get_local_size()