Does KNNSearch not support CUDA?

### Checklist

- [X] I have searched for [similar issues](https://github.com/isl-org/Open3D-ML/issues).
- [X] I have tested with the [latest development wheel](http://www.open3d.org/docs/latest/getting_started.html#development-version-pip).
- [X] I have checked the [release documentation](http://www.open3d.org/docs/release/) and the [latest documentation](http://www.open3d.org/docs/latest/) (for `main` branch).

### My Question

open3d==0.18, code
```
import torch
from open3d.ml.torch.layers import KNNSearch

# 假设已经有了数据和查询点，这里使用随机数据作为示例
points = torch.rand(100, 3, dtype=torch.float32).cuda() # 点云数据，假设已经在 GPU 上
queries = torch.rand(10, 3, dtype=torch.float32).cuda() # 查询点，假设已经在 GPU 上

# 创建 KNNSearch 层实例
knn_search = KNNSearch()

# 执行 k-NN 搜索
k = 5
neighbors_index, neighbors_distance = knn_search(points, queries=queries, k=k)

# `neighbors_index` 和 `neighbors_distance` 分别是查询点的最近邻索引和距离

```
result
```
/root/miniconda3/envs/gaussian_splatting/bin/python /workspace/gaussian/gaussian-splatting/utils/density.py
Traceback (most recent call last):
File "/workspace/gaussian/gaussian-splatting/utils/density.py", line 42, in
neighbors_index, neighbors_distance = knn_search(points=points, queries=queries, k=k)
File "/root/miniconda3/envs/gaussian_splatting/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/root/miniconda3/envs/gaussian_splatting/lib/python3.7/site-packages/open3d/ml/torch/python/layers/neighbor_search.py", line 382, in forward
index_dtype=self.index_dtype)
File "/root/miniconda3/envs/gaussian_splatting/lib/python3.7/site-packages/open3d/ml/torch/python/ops.py", line 642, in knn_search
return_distances=return_distances))
File "/root/miniconda3/envs/gaussian_splatting/lib/python3.7/site-packages/torch/_ops.py", line 143, in __call__
return self._op(*args, **kwargs or {})
RuntimeError: KnnSearch does not support CUDA

```

Looking forward your answer

Start with open3d.ml.torch.layers.KNNSearch and the neighbor_search.py and ops.py locations shown in the traceback. Reproduce the example with CUDA tensors, then determine whether CUDA support is intended; done should be a resolved behavior or a documented limitation with an appropriate test or documentation update.

Written by the indexing model from the issue text.