[Doc] add backend to doc string of initialize_model_parallel (vllm-project#22142)

Signed-off-by: Andy Xie <andy.xning@gmail.com>

Author: andyxning

vllm/distributed/parallel_state.py

@@ -1013,6 +1013,7 @@ def initialize_model_parallel(
             parallelism.
         pipeline_model_parallel_size: number of GPUs used for pipeline model
             parallelism.
+        backend: name of torch distributed communication backend.
 
     Let's say we have a total of 8 GPUs denoted by g0 ... g7 and we
     use 2 GPUs to parallelize the model tensor, and 4 GPUs to parallelize