[2/N][Feat] Add MC2 communication method for MoE layers (#2469)

### What this PR does / why we need it?
This method replaces the previous all-gather approach for small numbers
of tokens.

The key changes include:
- A new `AscendFusedMoE` layer that handles token splitting, local
computation, and final aggregation via all-gather.
- Logic in the model runner to dynamically select between the new MC2
method and the existing all-gather method based on the number of input
tokens.
- Sharding the MoE communication mask across tensor-parallel ranks.

### Does this PR introduce _any_ user-facing change?
None.

### How was this patch tested?
Test case fixed.


- vLLM version: v0.10.1.1
- vLLM main:
b00e69f8ca

---------

Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>

tests/ut/test_utils.py

@@ -289,13 +289,13 @@ class TestUtils(TestBase):
         # ascend custom op is not registered
         utils.register_ascend_customop()
         # should call register_oot three
-        self.assertEqual(mock_customop.register_oot.call_count, 8)
+        self.assertEqual(mock_customop.register_oot.call_count, 9)
         self.assertTrue(utils._ASCEND_CUSTOMOP_IS_REIGISTERED)
 
         # ascend custom op is already registered
         utils.register_ascend_customop()
         # should not register_oot again, thus only called three in this ut
-        self.assertEqual(mock_customop.register_oot.call_count, 8)
+        self.assertEqual(mock_customop.register_oot.call_count, 9)
 
 
 class TestProfileExecuteDuration(TestBase):