补齐 Autocal 测点质量与独立 trial 验证

按已确认窗口比较重复性、加载卸载差异和零载恢复，并逐 trial 留出验证。55 项定向测试、已录视频复核闭环及代码文档检查通过。

apps/autocal/config/locales/zh_CN.json

@@ -490,6 +490,40 @@
     "depth_missing": "本帧无有效深度／形变信号",
     "flow_missing": "本帧无有效位移信号"
   },
+  "quality": {
+    "open": "重复性与验证…",
+    "title": "测点质量分析",
+    "notice": "仅使用人工已确认的有效窗口；结果仍为初步分析。视频调试、人工稳定确认及 trial 留出验证均不能证明实物准确度或参考力同步。",
+    "repeatability": "重复性",
+    "repeatability_note": "相同分支、相同人工参考力下，先计算各 trial 的均值，再计算 trial 间极差和样本标准差。至少两次 trial 才有重复性统计；这不同于保持段内波动。",
+    "hysteresis": "加载／卸载差异",
+    "hysteresis_note": "只配对同一 trial 内人工参考力完全相同的非零加载与卸载窗口，不插值。差值为卸载减加载，单位 raw；缺少匹配档位时不生成结果。",
+    "zero_return": "零载恢复",
+    "zero_return_note": "同一 trial 的最终卸载零载减初始零载，单位 raw；只配对人工参考力为 0 N 的已确认窗口。",
+    "validation": "独立 trial 验证",
+    "validation_note": "逐次留出一个完整 trial，用其余 trial 的零载／加载窗口拟合，再预测留出 trial 的全部已确认窗口。每窗口等权；这是交叉验证，不能代替另一场独立实验。",
+    "validation_summary": "验证窗口 {count} 个；RMSE {rmse} N；MAE {mae} N；最大绝对误差 {maximum} N；超出训练 raw 范围 {outside} 个。",
+    "no_data": "尚无可用数据，请先复核所需窗口。",
+    "insufficient_training": "其余 trial 的训练数据不足",
+    "outside_range": "超出训练 raw 范围",
+    "in_range": "训练 raw 范围内",
+    "col_branch": "阶段",
+    "col_load": "人工参考力 (N)",
+    "col_trials": "trial 数",
+    "col_range": "trial 间极差 (raw)",
+    "col_std": "trial 间样本标准差 (raw)",
+    "col_trial": "trial",
+    "col_load_raw": "加载均值 (raw)",
+    "col_unload_raw": "卸载均值 (raw)",
+    "col_delta_raw": "差值 (raw)",
+    "col_initial_zero": "初始零载 (raw)",
+    "col_final_zero": "最终零载 (raw)",
+    "col_held_out": "留出验证 trial",
+    "col_training": "训练 trial",
+    "col_predicted": "预测力 (N)",
+    "col_error": "预测减参考 (N)",
+    "col_range_status": "范围／状态"
+  },
   "review": {
     "load_selection": "载入已保存复核…",
     "confirm_stable": "已检查此段稳定",