IndexError during inference

### 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).

### Describe the issue

I have trained RandlaNet architecture on SensatUrban dataset by creating a dedicated dataset class using toronto3d template. I have trained the architecture with a grid size of 0.05 but when I try to perform an inference on the test set I get IndexError in the update_tests function. I think it might be related to the grid size but I can not understand why.

### Steps to reproduce the bug

```python
Config file

dataset:
name: SensatUrban
cache_dir: ./dataset_cache
dataset_path: data/digitwin/sensatUrban/ply
class_weights: [438547425, 543847609, 860487257, 20606217, 3044807, 49452238, 498977, 131470199, 26669467, 37557130, 43503086, 174125, 7088841]
num_classes: 13
num_points: 65536
train_files:
- train/birmingham_block_10.ply
- train/birmingham_block_11.ply
- train/birmingham_block_12.ply
- train/birmingham_block_13.ply
- train/birmingham_block_3.ply
- train/birmingham_block_4.ply
- train/birmingham_block_6.ply
- train/cambridge_block_12.ply
- train/cambridge_block_13.ply
- train/cambridge_block_14.ply
- train/cambridge_block_17.ply
- train/cambridge_block_18.ply
- train/cambridge_block_19.ply
- train/cambridge_block_2.ply
- train/cambridge_block_20.ply
- train/cambridge_block_21.ply
- train/cambridge_block_23.ply
- train/cambridge_block_25.ply
- train/cambridge_block_26.ply
- train/cambridge_block_28.ply
- train/cambridge_block_3.ply
- train/cambridge_block_32.ply
- train/cambridge_block_33.ply
- train/cambridge_block_34.ply
- train/cambridge_block_4.ply
- train/cambridge_block_6.ply
- train/birmingham_block_0.ply
- train/birmingham_block_7.ply
- train/birmingham_block_9.ply
- train/cambridge_block_0.ply
- train/cambridge_block_1.ply
- train/cambridge_block_8.ply
- train/cambridge_block_9.ply
val_files:
- train/birmingham_block_1.ply
- train/birmingham_block_5.ply
- train/cambridge_block_10.ply
- train/cambridge_block_7.ply
test_files:
- test/birmingham_block_2.ply
- test/birmingham_block_8.ply
- test/cambridge_block_15.ply
- test/cambridge_block_22.ply
- test/cambridge_block_16.ply
- test/cambridge_block_27.ply
test_result_folder: ./test
use_cache: true
steps_per_epoch_train: 100
steps_per_epoch_valid: 10
use_cache: true
model:
name: RandLANet
batcher: DefaultBatcher
ckpt_path: # path/to/your/checkpoint
num_neighbors: 16
num_layers: 5
num_points: 65536
num_classes: 13
ignored_label_inds: []
sub_sampling_ratio: [4, 4, 4, 4, 2]
in_channels: 6
dim_input: 6
dim_features: 8
dim_output: [16, 64, 128, 256, 512]
grid_size: 0.05
augment:
recenter:
dim: [0, 1, 2]
normalize:
points:
method: linear
pipeline:
name: SemanticSegmentation
optimizer:
lr: 0.01
batch_size: 4
main_log_dir: experiments/sensat
max_epoch: 800
save_ckpt_freq: 1
scheduler_gamma: 0.99
test_batch_size: 1
num_workers: 0
pin_memory: false
val_batch_size: 2
experiment: randlanet_sensat_weight_paper
summary:
record_for: []
max_pts:
use_reference: false
max_outputs: 1
```
python Open3D-ML/scripts/run_pipeline.py torch \
--split test \
-c {CFG_PATH} \
--model.ckpt_path {MODEL_PATH} \
--dataset.test_result_folder {OUT_PATH}
```

### Error message

```python
Traceback (most recent call last):
File "Open3D-ML/scripts/run_pipeline.py", line 301, in
sys.exit(main())
File "Open3D-ML/scripts/run_pipeline.py", line 213, in main
pipeline.run_test()
File "/nfs/home/blanco/digitwin/Open3D-ML/ml3d/torch/pipelines/semantic_segmentation.py", line 242, in run_test
self.update_tests(test_sampler, inputs, results)
File "/nfs/home/blanco/digitwin/Open3D-ML/ml3d/torch/pipelines/semantic_segmentation.py", line 323, in update_tests
test_labels = np.argmax(self.test_probs[self.curr_cloud_id][proj_inds], 1)
IndexError: index 10379412 is out of bounds for axis 0 with size 3444876
```

### Expected behavior

Since the training completes smoothly I would expect the inference will go smoothly as well.

### Open3D, Python and System information

```markdown
- Operating system: Ubuntu 22.04.4 LTS
- Python version: Python 3.8.5
- Open3D version: 0.16.0
- System type: x86_64
- Is this remote workstation?: yes
- How did you install Open3D?: pip
```

### Additional information

_No response_