feat(scripts): 数据转换、ONNX 导出与推理性能分析

- convert_annotations: labelme / VOC -> YOLO txt，顺带生成 data.yaml
- split_dataset: 切分成 images/{train,val} 与 labels/{train,val}，默认建软链不复制数据
- export_onnx: 导出部署用 ONNX，CPU 可跑，缺依赖时提示 `uv sync --group export`
- profile_inference: torch.profiler 测单帧延迟/FPS、CUDA 显存峰值与算子级明细
- check_env: Python / PyTorch / Nvidia GPU 自检

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

scripts/check_env.py

@@ -0,0 +1,41 @@
+#!/usr/bin/env python
+"""环境自检：确认依赖与 Nvidia GPU 是否就绪。
+
+python scripts/check_env.py
+"""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+from utils.device import NoNvidiaGpuError, cuda_devices, require_cuda  # noqa: E402
+
+
+def main() -> int:
+    print(f"Python: {sys.version.split()[0]}")
+    try:
+        import torch
+
+        print(f"PyTorch: {torch.__version__} (CUDA build: {torch.version.cuda})")
+    except ImportError:
+        print("PyTorch: 未安装 —— 先执行 `uv sync`")
+        return 1
+
+    devices = cuda_devices()
+    print(f"Nvidia GPU: {devices if devices else '无'}")
+
+    try:
+        require_cuda()
+    except NoNvidiaGpuError as exc:
+        print(f"训练不可用: {exc}")
+        print("（评估/导出仍可在 CPU 上进行）")
+        return 1
+    print("训练环境就绪。")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())