[Feat][SP] Suport SP for VL MoE models (#7044)

### What this PR does / why we need it?

2nd PR for https://github.com/vllm-project/vllm-ascend/issues/5712,
extend SP to VL MoE models.


### Does this PR introduce _any_ user-facing change?
remove `sp_threshold` in additional config and reuse `sp_min_token_num`
from vLLM.


### How was this patch tested?
- Model: Qwen3-VL-30B-A3B, 
- TP4 DP2
- 100 reqs
- max concurrency 1

| Seq length | Mean TTFT (ms) main | Mean TTFT (ms) this PR |
|------------|---------------------|------------------------|
| 4k         | 429.40               | 323.3                  |
| 16k        | 1297.01              | 911.74                |

- vLLM version: v0.16.0
- vLLM main:
4034c3d32e

---------

Signed-off-by: realliujiaxu <realliujiaxu@163.com>

tests/e2e/singlecard/compile/backend.py

@@ -100,6 +100,11 @@ class TestBackend:
         """Helper to find all FX nodes that call a specific operator."""
         return [node for node in graph.graph.nodes if hasattr(node, "target") and node.target == target]
 
+    def op_count(self, op: OpOverload, before: bool = False) -> int:
+        """Return the number of nodes that call the given operator."""
+        graph = self.graph_pre_pass if before else self.graph_post_pass
+        return len(self.find_nodes_by_target(graph, op))
+
     def check_before_ops(self, ops: Sequence[OpOverload], fully_replaced: bool = True):
         """
         Verify that the original (unfused) operators exist before the pass