fix(networks): replace Tensor | None union syntax with Optional[Tensor] for TorchScript compatibility

The `|` union type syntax (e.g. `torch.Tensor | None`) was introduced in
Python 3.10. While `from __future__ import annotations` defers evaluation
at runtime, TorchScript's annotation parser does not support this syntax
and fails when scripting models that contain these forward method signatures.

Replace `torch.Tensor | None` with `Optional[torch.Tensor]` in the
`forward` methods of:
- `monai/networks/blocks/crossattention.py` (CrossAttentionBlock)
- `monai/networks/blocks/selfattention.py` (SABlock)
- `monai/networks/blocks/transformerblock.py` (TransformerBlock)

These three blocks are used in the ViT/UNETR scripting path, causing
`RuntimeError: Can't redefine method: forward on class` when
`torch.jit.script()` is called on a UNETR model.

Closes #7939

Signed-off-by: Oleksandr Sanin <alexaaander.sanin@gmail.com>

Author: AlexanderSanin

monai/networks/blocks/crossattention.py

@@ -13,6 +13,7 @@
 
 import torch
 import torch.nn as nn
+from typing import Optional
 
 from monai.networks.layers.utils import get_rel_pos_embedding_layer
 from monai.utils import optional_import
@@ -139,7 +140,7 @@ def __init__(
         )
         self.input_size = input_size
 
-    def forward(self, x: torch.Tensor, context: torch.Tensor | None = None):
+    def forward(self, x: torch.Tensor, context: Optional[torch.Tensor] = None):
         """
         Args:
             x (torch.Tensor): input tensor. B x (s_dim_1 * ... * s_dim_n) x C

monai/networks/blocks/selfattention.py

@@ -14,6 +14,7 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
+from typing import Optional
 
 from monai.networks.layers.utils import get_rel_pos_embedding_layer
 from monai.utils import optional_import
@@ -158,7 +159,7 @@ def __init__(
         )
         self.input_size = input_size
 
-    def forward(self, x, attn_mask: torch.Tensor | None = None):
+    def forward(self, x, attn_mask: Optional[torch.Tensor] = None):
         """
         Args:
             x (torch.Tensor): input tensor. B x (s_dim_1 * ... * s_dim_n) x C

monai/networks/blocks/transformerblock.py

@@ -13,6 +13,7 @@
 
 import torch
 import torch.nn as nn
+from typing import Optional
 
 from monai.networks.blocks import CrossAttentionBlock, MLPBlock, SABlock
 
@@ -89,7 +90,7 @@ def __init__(
         )
 
     def forward(
-        self, x: torch.Tensor, context: torch.Tensor | None = None, attn_mask: torch.Tensor | None = None
+        self, x: torch.Tensor, context: Optional[torch.Tensor] = None, attn_mask: Optional[torch.Tensor] = None
     ) -> torch.Tensor:
         x = x + self.attn(self.norm1(x), attn_mask=attn_mask)
         if self.with_cross_attention: