[Perf][Diffusion] Optimize HSDP startup via Rank-0 shared weight loading and accelerated LoRA delta computation (#7005)

Signed-off-by: samithuang <285365963@qq.com>

vllm_omni/diffusion/data.py

@@ -920,6 +920,7 @@ class OmniDiffusionConfig:
 
     # Parallel weight loading (for faster diffusion model startup)
     enable_multithread_weight_load: bool = True
+    enable_broadcast_weight_load: bool = False
     num_weight_load_threads: int = 4
 
     # Enable sleep mode