revert: 回退 StandardScaler — 跨传感器场景下双重归一化导致推理坍缩

- 删除 Pipeline 中 scaler(StandardScaler) 步骤
- 删除 save_model 中 scaler_mean/scaler_scale 导出
- 删除 StandardScaler import
- 恢复: imputer → FeatureUnion → cleaner → SVR
- MNF 白化本身就是尺度对齐，无需额外标准化

src/core/modeling/modeling_batch.py

@@ -10,7 +10,6 @@ warnings.filterwarnings('ignore')
 
 # 机器学习模型导入 - 改为回归模型
 from sklearn.base import BaseEstimator, TransformerMixin
-from sklearn.preprocessing import StandardScaler
 from sklearn.svm import SVR
 from sklearn.ensemble import RandomForestRegressor
 from sklearn.neighbors import KNeighborsRegressor
@@ -647,7 +646,6 @@ class WaterQualityModelingBatch:
             ('imputer', SimpleImputer(strategy='median')),
             ('preproc', preproc),
             ('cleaner', _SafeFiniteTransformer()),
-            ('scaler',  StandardScaler()),
             ('model',   base_model),
         ])
 
@@ -787,11 +785,6 @@ class WaterQualityModelingBatch:
                 'mnf_W':    np.asarray(_mnf.W_mnf_),
                 'mnf_n_components': int(getattr(_mnf, 'n_components_', _mnf.n_components)),
             }
-            # StandardScaler 参数（C++ 端需复现 (X - mean) / scale）
-            _scaler = model.named_steps.get('scaler')
-            if _scaler is not None and hasattr(_scaler, 'mean_'):
-                _deploy['scaler_mean']  = np.asarray(_scaler.mean_, dtype=np.float64)
-                _deploy['scaler_scale'] = np.asarray(_scaler.scale_, dtype=np.float64)
             # SVR 部署矩阵（rbf kernel 需要 support_vectors_）
             _deploy['svr_dual_coef'] = np.asarray(_svr.dual_coef_)
             _deploy['svr_intercept'] = np.asarray(_svr.intercept_)