[BugFix] Fix invalid flashinfer_fused_moe_blockscale_fp8 op registration (#32855)

Signed-off-by: Fadi Arafeh <fadi.arafeh@arm.com>

vllm/model_executor/layers/fused_moe/flashinfer_trtllm_moe.py

@@ -131,7 +131,7 @@ def flashinfer_fused_moe_blockscale_fp8(
     expert_offset: int,
     local_num_experts: int,
     block_shape: list[int],
-    routing_method_type: int = RoutingMethodType.DeepSeekV3,
+    routing_method_type: int = int(RoutingMethodType.DeepSeekV3),
     routed_scaling: float | None = 1.0,
 ) -> torch.Tensor:
     from vllm.utils.flashinfer import flashinfer_trtllm_fp8_block_scale_moe