fix: load model with explicit weights_only=False

Torch 2.4.0 shows a FutureWarning about the default value of this arg changing if it's not set explicitly.

src/anemoi/inference/runner.py

@@ -266,10 +266,7 @@ def run(
 
         with Timer(f"Loading {self.checkpoint}"):
             try:
-                model = torch.load(
-                    self.checkpoint.path,
-                    map_location=device,
-                ).to(device)
+                model = torch.load(self.checkpoint.path, map_location=device, weights_only=False).to(device)
             except Exception:
                 self.checkpoint.report_loading_error()
                 raise