[Feat] Dynamic Batch Feature (#3490)

[RFC](https://github.com/vllm-project/vllm-ascend/issues/3328) for more
details.
Add dynamic batch feature in chunked prefilling strategy, the token
budget can be refined to achieve better effective throughput and TPOT.

!!! NOTE: only 910B3 is supported till now, we are working on further
improvements.
Additional file for lookup table is required.

- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: Cheng Wang <wangchengkyrie@outlook.com>

vllm_ascend/platform.py

@@ -314,6 +314,14 @@ class NPUPlatform(Platform):
                 vllm_config.scheduler_config)
             vllm_config.scheduler_config = recompute_scheduler_config
 
+        # Extend original scheduler_config to use SchedulerDynamicBatch.
+        if ascend_config.SLO_limits_for_dynamic_batch != -1:
+            vllm_config.scheduler_config.scheduler_cls = (
+                "vllm_ascend.core.scheduler_dynamic_batch.SchedulerDynamicBatch"
+            )
+            vllm_config.scheduler_config.chunked_prefill_enabled = True
+            vllm_config.scheduler_config.SLO_limits_for_dynamic_batch = ascend_config.SLO_limits_for_dynamic_batch
+
     @classmethod
     def get_attn_backend_cls(
         cls,