[main][prefill optimization] Optimize parallel strategies to reduce communication overhead (#2198)

### What this PR does / why we need it?
1.Shared Expert Sharding Strategy Update: Switched from TP-aligned to
pure DP for shared experts, enabling more efficient execution.
2.O_Proj AllReduce → ReduceScatter: Reduced communication overhead by
using ReduceScatter, made possible by pure DP sharding.
3.AllGather Postponed: Delayed to after QKV down projection to reduce
synchronization impact during prefill.

### How was this patch tested?
Adding ut case in `tests/ut/attention/test_mla_v1.py`

#### How to run

use parameter `--additional_config='{"enable_shared_expert_dp": true}'`

##### a.How to run eager mode

eg:
python -m vllm.entrypoints.openai.api_server --model=/model_path
--trust-remote-code -tp 8 -dp 2 --enable_expert_parallel --port 8002
--max-model-len 5120 --max-num-batched-tokens 16384 --enforce-eager
--disable-log-requests
--additional_config='{"ascend_scheduler_config":{"enabled":true},"enable_shared_expert_dp":
true,"chunked_prefill_for_mla":true}'

##### b.How to run graph mode

eg:
python -m vllm.entrypoints.openai.api_server --model=/model_path
--trust-remote-code -tp 8 -dp 2 --enable_expert_parallel --port 8002
--max-model-len 5120 --max-num-batched-tokens 16384
--disable-log-requests
--additional_config='{"ascend_scheduler_config":{"enabled":true},"enable_shared_expert_dp":
true,"chunked_prefill_for_mla":true,"torchair_graph_config":{"enabled":true}}'


- vLLM version: v0.10.0
- vLLM main:
9edd1db02b

---------

Signed-off-by: Wang Kunpeng <1289706727@qq.com>
Signed-off-by: SlightwindSec <slightwindsec@gmail.com>
Co-authored-by: SlightwindSec <slightwindsec@gmail.com>

vllm_ascend/ascend_config.py

@@ -47,6 +47,9 @@ class AscendConfig:
         self.expert_map_path = additional_config.get("expert_map_path", None)
         self.chunked_prefill_for_mla = additional_config.get(
             "chunked_prefill_for_mla", False)
+        self.enable_shared_expert_dp = additional_config.get(
+            "enable_shared_expert_dp", True
+        ) and not self.torchair_graph_config.enabled and vllm_config.parallel_config.enable_expert_parallel
 
 
 class TorchairGraphConfig:
@@ -166,6 +169,10 @@ def check_ascend_config(vllm_config, enforce_eager):
                     raise NotImplementedError(
                         "Torchair graph mode only works with following model types:"
                         f"{TORCHAIR_MODEL_LIST}.")
+            if ascend_config.enable_shared_expert_dp:
+                logger.warning(
+                    "enable_shared_expert_dp is not supported for torchair graph mode currently, "
+                    "it has been disabled automatically.")
         # aclgraph case
         else:
             # aclgraph doesn't work with deepseek model and only qwen model is well tested.