Fix error message typo and remove redundant type casts (#3503)

Addresses code review findings and Pyrefly warnings in metrics module.

## Error Message Correction
`MeanPairwiseDistance.compute()` raised `NotComputableError` with
incorrect class name reference:

```python
# Before
raise NotComputableError("MeanAbsoluteError must have at least one example...")

# After
raise NotComputableError("MeanPairwiseDistance must have at least one example...")
```

Updated corresponding test expectation in
`test_mean_pairwise_distance.py::test_zero_sample`.

## Redundant Cast Removal
Removed unnecessary `cast()` calls in `Loss.update()` where function
signature already provides type narrowing:

```python
def update(self, output: tuple[Tensor, Tensor] | tuple[Tensor, Tensor, dict]) -> None:
    if len(output) == 2:
        y_pred, y = output  # cast() removed - type already constrained
    else:
        y_pred, y, kwargs = output  # cast() removed - type already constrained
```

Removed unused `cast` import from `typing`.

<!-- START COPILOT CODING AGENT TIPS -->
---

💬 We'd love your input! Share your thoughts on Copilot coding agent in
our [2 minute survey](https://gh.io/copilot-coding-agent-survey).

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: vfdev-5 <2459423+vfdev-5@users.noreply.github.com>

Author: Copilot

ignite/metrics/loss.py

@@ -1,4 +1,4 @@
-from typing import Callable, cast
+from typing import Callable
 
 import torch
 
@@ -92,10 +92,10 @@ def reset(self) -> None:
     @reinit__is_reduced
     def update(self, output: tuple[torch.Tensor, torch.Tensor] | tuple[torch.Tensor, torch.Tensor, dict]) -> None:
         if len(output) == 2:
-            y_pred, y = cast(tuple[torch.Tensor, torch.Tensor], output)
+            y_pred, y = output
             kwargs: dict = {}
         else:
-            y_pred, y, kwargs = cast(tuple[torch.Tensor, torch.Tensor, dict], output)
+            y_pred, y, kwargs = output
         average_loss = self._loss_fn(y_pred, y, **kwargs).detach()
 
         if len(average_loss.shape) != 0:

ignite/metrics/mean_pairwise_distance.py

@@ -1,4 +1,4 @@
-from typing import Callable, Sequence, Union
+from typing import Callable, Sequence
 
 import torch
 from torch.nn.functional import pairwise_distance
@@ -72,7 +72,7 @@ def __init__(
         p: int = 2,
         eps: float = 1e-6,
         output_transform: Callable = lambda x: x,
-        device: Union[str, torch.device] = torch.device("cpu"),
+        device: str | torch.device = torch.device("cpu"),
         skip_unrolling: bool = False,
     ) -> None:
         super(MeanPairwiseDistance, self).__init__(output_transform, device=device, skip_unrolling=False)
@@ -92,7 +92,7 @@ def update(self, output: Sequence[torch.Tensor]) -> None:
         self._num_examples += y.shape[0]
 
     @sync_all_reduce("_sum_of_distances", "_num_examples")
-    def compute(self) -> Union[float, torch.Tensor]:
+    def compute(self) -> float | torch.Tensor:
         if self._num_examples == 0:
-            raise NotComputableError("MeanAbsoluteError must have at least one example before it can be computed.")
+            raise NotComputableError("MeanPairwiseDistance must have at least one example before it can be computed.")
         return self._sum_of_distances.item() / self._num_examples

tests/ignite/metrics/test_mean_pairwise_distance.py

@@ -12,7 +12,7 @@
 def test_zero_sample():
     mpd = MeanPairwiseDistance()
     with pytest.raises(
-        NotComputableError, match=r"MeanAbsoluteError must have at least one example before it can be computed"
+        NotComputableError, match=r"MeanPairwiseDistance must have at least one example before it can be computed"
     ):
         mpd.compute()