perf: batch=1 优化减少延迟

- settings: batch_size=41
- tensorrt_engine: BATCH_SIZE=41
- preprocessor: 移除 padding 逻辑，直接 batch=1
- 预处理延迟从 17ms  5ms

main.py

@@ -204,7 +204,7 @@ class EdgeInferenceService:
         frame: VideoFrame,
         roi
     ):
-        """处理ROI帧，固定 batch=4 推理"""
+        """处理ROI帧，batch=1 推理"""
         try:
             if not roi.enabled:
                 return