use dist.all_reduce(min) to sync num_blocks_local (#2933)

* pre-commit all files check

* reduce min num_blocks_local

* fix nranks=1

* pre-commit when commit-msg

fastdeploy/model_executor/forward_meta.py

@@ -110,6 +110,7 @@ class XPUForwardMeta(ForwardMeta):
     """
     XPUForwardMeta is used to store the global meta information of the forward, and some XPU specific meta info.
     """
+
     # Accumulated offset
     cum_offsets: Optional[paddle.Tensor] = None
     # TODO(wanghaitao): Supplementary notes