Fix GridSampling3D fails when called with a cuda data tensor without batch key

[fifarraj]

Currently when GridSampling3D is called with a cuda data tensor which does not have a "batch" key, it fails with the below exception. This PR fixes it.

Exception has occurred: RuntimeError       (note: full exception trace is shown but execution is paused at: _run_module_as_main)
The following operation failed in the TorchScript interpreter.
Traceback of TorchScript (most recent call last):
  File "/home/firas/miniconda3/envs/legged_gym_crash_torch_121/lib/python3.8/site-packages/torch_cluster/grid.py", line 32, in grid_cluster
        cluster = grid_cluster(pos, size)
    """
    return torch.ops.torch_cluster.grid(pos, size, start, end)
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <--- HERE
RuntimeError: size.device().is_cuda() INTERNAL ASSERT FAILED at "csrc/cuda/grid_cuda.cu":31, please report a bug to PyTorch. size must be CUDA tensor


  File "torch_points3d/core/data_transform/grid_transform.py", line 119, in _process
    cluster = grid_cluster(coords, torch.tensor([1, 1, 1]))
  File "torch_points3d/core/data_transform/grid_transform.py", line 135, in __call__
    data = self._process(data)

To test:

import torch
from torch_points3d.core.data_transform import GridSampling3D
from torch_geometric.data import Batch

data = torch.tensor([[0,0,0]], device='cuda')
data = Batch(pos=data)
pcd_to_grid = GridSampling3D(torch.tensor([0.2], device='cuda'))
pcd_to_grid(data)