fix(runner): 14 Facade kwargs 兜底 + 4 spec parameter_map 修正 + step6_75 路由切到 indices

- 14 个 stepX_... Facade 形参表末尾加 **kwargs，杜绝 Runner 注入未声明 key 时的 TypeError（典型：step3 收到 glint_mask_path）

- runner._invoke user_overrides 合并加 v is not None and v != '' 过滤，避免 GUI 面板空值覆盖 ctx 中已写入的有效路径

- PIPELINE_STEPS 加 4 个 parameter_map 修正 ctx 字段名→形参名错位：step6_5/6_75: training_csv_path→csv_path；step8_5: models_dir→non_empirical_models_dir；step8_75: models_dir→custom_regression_dir

- step6_75 路由从 training_csv_path 切到 indices_path（requires + parameter_map 同步）；配合 skip_when_missing，未跑 step5_5 时自动 skip

- worker_thread.py: mode='full' 切到 PipelineRunner + PipelineContext 调度

src/core/pipeline/runner.py

@@ -84,11 +84,13 @@ PIPELINE_STEPS: List[StepSpec] = [
     StepSpec(
         step_id="step6_5", method_name="step6_5_non_empirical_modeling",
         requires=["training_csv_path"], produces=["models_dir"],
+        parameter_map={"training_csv_path": "csv_path"},
         description="非经验统计回归",
     ),
     StepSpec(
         step_id="step6_75", method_name="step6_75_custom_regression",
-        requires=["training_csv_path"], produces=["models_dir"],
+        requires=["indices_path"], produces=["models_dir"],
+        parameter_map={"indices_path": "csv_path"},
         description="自定义回归分析",
     ),
     StepSpec(
@@ -104,12 +106,14 @@ PIPELINE_STEPS: List[StepSpec] = [
     StepSpec(
         step_id="step8_5", method_name="step8_5_predict_with_non_empirical_models",
         requires=["sampling_csv_path", "models_dir"], produces=["prediction_dir"],
+        parameter_map={"models_dir": "non_empirical_models_dir"},
         description="非经验模型预测",
     ),
     StepSpec(
         step_id="step8_75", method_name="step8_75_predict_with_custom_regression",
         requires=["sampling_csv_path", "models_dir", "formula_csv_path"],
         produces=["prediction_dir"],
+        parameter_map={"models_dir": "custom_regression_dir"},
         description="自定义回归预测",
     ),
     StepSpec(
@@ -185,7 +189,10 @@ class PipelineRunner:
         # 2) 允许用户在 ctx.user_config[step_id] 覆盖/补充
         user_overrides = ctx.user_config.get(spec.step_id) or {}
         if isinstance(user_overrides, dict):
-            kwargs.update(user_overrides)
+            for k, v in user_overrides.items():
+                # ★ 关键防御：绝不用 GUI 的“空字符串”或 None 覆盖上游传来的有效路径
+                if v is not None and v != "":
+                    kwargs[k] = v
 
         # 3) 状态置 start
         ctx.append_log(