Allow AsyncLLMEngine.generate to target a specific DP rank (#19102)

Signed-off-by: Jon Swenson <jmswen@gmail.com>

vllm/v1/engine/processor.py

@@ -212,6 +212,7 @@ class Processor:
         trace_headers: Optional[Mapping[str, str]] = None,
         prompt_adapter_request: Optional[PromptAdapterRequest] = None,
         priority: int = 0,
+        data_parallel_rank: Optional[int] = None,
     ) -> tuple[Optional[str], EngineCoreRequest]:
 
         # TODO(woosuk): Support pooling models.
@@ -328,6 +329,7 @@ class Processor:
             arrival_time=arrival_time,
             lora_request=lora_request,
             cache_salt=decoder_inputs.get("cache_salt"),
+            data_parallel_rank=data_parallel_rank,
         )
 
     def _validate_model_inputs(self,