feat: 接入 SCDM 优先编辑闭环并支持阵列局部间距

接入 SCDM probe/edit/cache/校验链路，增强孔组、阵列、关系式和参数表交互。

支持阵列相邻段间距、移动意图切换、结果回滚校验，并补充对应回归脚本。

step_editor/relation_formulas.py

@@ -144,6 +144,44 @@ def evaluate_relation_formula(
     return _coerce_formula_value(value)
 
 
+def validate_relation_formula_graph(formulas: Iterable[RelationFormula]) -> None:
+    target_to_formula: dict[str, RelationFormula] = {}
+    for formula in formulas:
+        target_token = formula.target.token
+        if target_token in target_to_formula:
+            raise RelationFormulaError(f"同一目标参数只能由一条关系式控制：{target_token}。")
+        target_to_formula[target_token] = formula
+
+    target_tokens = set(target_to_formula)
+    graph: dict[str, list[str]] = {}
+    for target_token, formula in target_to_formula.items():
+        reference_tokens = [ref.token for ref in formula.references]
+        if target_token in reference_tokens:
+            raise RelationFormulaError(f"关系式不能引用自身：{target_token}。")
+        graph[target_token] = [ref_token for ref_token in reference_tokens if ref_token in target_tokens]
+
+    visit_state: dict[str, str] = {}
+    stack: list[str] = []
+
+    def visit(token: str) -> None:
+        state = visit_state.get(token)
+        if state == "visiting":
+            start_index = stack.index(token) if token in stack else 0
+            cycle = [*stack[start_index:], token]
+            raise RelationFormulaError(f"关系式存在循环依赖：{' -> '.join(cycle)}。")
+        if state == "visited":
+            return
+        visit_state[token] = "visiting"
+        stack.append(token)
+        for dependency in graph.get(token, []):
+            visit(dependency)
+        stack.pop()
+        visit_state[token] = "visited"
+
+    for target_token in graph:
+        visit(target_token)
+
+
 def relation_value_to_text(value: object) -> str:
     value = _coerce_formula_value(value)
     if isinstance(value, Vector3):