fix enseble data typing due to pytorch update

src/secmlt/adv/evasion/aggregators/ensemble.py

@@ -18,7 +18,7 @@ def __call__(
         model: BaseModel,
         data_loader: DataLoader,
         adv_loaders: list[DataLoader],
-    ) -> DataLoader[torch.Tuple[torch.Tensor]]:
+    ) -> DataLoader[tuple[torch.Tensor]]:
         """
         Get the worst-case of the metric with the given implemented criterion.