Fix AttributeError in torch_labels_to_contours when torch is not installed

- Use string literals for torch.Tensor type hints to prevent AttributeError
- Add TYPE_CHECKING import for proper type hint handling
- Add from __future__ import annotations for forward references

Author: macromeer

src/napari_tmidas/processing_functions/torch_labels_to_contours.py

@@ -10,23 +10,28 @@
 PyTorch's GPU acceleration instead of CuPy.
 """
 
+from __future__ import annotations
+
 from pathlib import Path
-from typing import List, Optional, Sequence, Tuple, Union
+from typing import TYPE_CHECKING, List, Optional, Sequence, Tuple, Union
 
 import numpy as np
 import zarr
 from tqdm import tqdm
 
+if TYPE_CHECKING:
+    import torch
+
 try:
     import torch
     import torch.nn.functional as F
     TORCH_AVAILABLE = True
 except ImportError:
     TORCH_AVAILABLE = False
-    torch = None
+    torch = None  # type: ignore
 
 
-def _find_boundaries_torch(labels: torch.Tensor, mode: str = "outer") -> torch.Tensor:
+def _find_boundaries_torch(labels: "torch.Tensor", mode: str = "outer") -> "torch.Tensor":
     """
     PyTorch implementation of skimage.segmentation.find_boundaries.
     
@@ -113,7 +118,7 @@ def _find_boundaries_torch(labels: torch.Tensor, mode: str = "outer") -> torch.T
     return boundary
 
 
-def _gaussian_filter_torch(input: torch.Tensor, sigma: float) -> torch.Tensor:
+def _gaussian_filter_torch(input: "torch.Tensor", sigma: float) -> "torch.Tensor":
     """
     PyTorch implementation of Gaussian filtering.