refactor(cdsl_engine): executor registry + per-family executor package

Phase 3 of the decoupling refactor (behavior-preserving):
- registry.py: atomic_executor decorator, ALL_ATOMIC_IDS with
  fail-fast registration validation, execute_node dispatcher
- executors/: one module per family (extrude, revolve, surfaces,
  loft_sweep, bodies, context, primitives, parametric, holes,
  dressup, patterns) + shared helpers in executors/common
- executors/__init__: explicit aggregation + completeness check
  (registry must cover every declared atomic id at import time)
- runtime.py: slimmed to entry points (analyze_cdsl/rebuild_cdsl)
  plus full historical re-exports incl. test-referenced privates

Adding an atomic operation now touches only one executor module and
its schema contract; the shared registry never changes. Verified
against baseline: zero new failures.

backend/engine/cdsl_engine/executors/context.py

@@ -0,0 +1,63 @@
+"""Reference-geometry executors (reference_plane / reference_axis).
+
+Context features produce no solid; they register durable topology contexts
+that later features resolve through owner-qualified selectors.
+"""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any
+
+from ..registry import atomic_executor
+from ..specs import AxisSpec, PlaneSpec, vector_add, vector_cross, vector_dot, vector_scale, vector_unit
+from ..topology import FeaturePlanNode, FeatureResult
+
+if TYPE_CHECKING:  # pragma: no cover - import for type checkers only
+    from ..session import ExecutionSession
+
+
+@atomic_executor("reference_plane")
+def _reference_plane_executor(node: FeaturePlanNode, session: "ExecutionSession", sketch: dict[str, Any] | None) -> FeatureResult:
+    del sketch
+    # 基准面特征（reference_plane）执行入口：从参数解析平面并登记为拓扑上下文。
+
+    # 1. 从特征参数 plane 中解析出平面定义 PlaneSpec（原点到法向）。
+    plane = PlaneSpec.from_mapping(node.params.get("plane") or {})
+    # 2. 将该平面注册到拓扑上下文，供后续特征（如草图基准、参考轴）引用。
+    session.topology.register_context(node.feature_id, plane)
+    # 3. 返回结果对象，并将该平面作为上下文一并携带。
+    return session.result(node, context=plane)
+
+
+@atomic_executor("reference_axis")
+def _reference_axis_executor(node: FeaturePlanNode, session: "ExecutionSession", sketch: dict[str, Any] | None) -> FeatureResult:
+    del sketch
+    # 基准轴特征（reference_axis）执行入口：由参数直接定义轴，或由两个基准平面求交线得到轴。
+
+    # 1. 尝试直接取参数：若同时给出原点 origin_mm 与方向 direction，则直接构造轴。
+    params = node.params.get("axis") or {}
+    if params.get("origin_mm") and params.get("direction"):
+        axis = AxisSpec.from_mapping(params)
+    else:
+        # 2. 否则从特征选择器中筛选出已解析的基准平面。
+        planes = [session.resolve(selector) for selector in node.selectors if selector.get("kind") == "plane"]
+        resolved = [item.record.value for item in planes if item.status == "resolved" and isinstance(item.record.value, PlaneSpec)]
+        # 3. 校验：轴需要两个非平行的平面，不足两个则报错。
+        if len(resolved) < 2:
+            raise ValueError("reference axis requires two uniquely resolved planes")
+        # 4. 用两平面法线叉积求交线方向；若方向长度接近 0 说明两平面平行，无法成轴。
+        first, second = resolved[0], resolved[1]
+        n1, n2 = first.normal, second.normal
+        direction = vector_cross(n1, n2)
+        squared_length = vector_dot(direction, direction)
+        if squared_length <= 1e-18:
+            raise ValueError("reference planes are parallel and cannot define an axis")
+        # 5. 求交线上的一点：两平面到各自原点的垂距参与线性组合，得到交线上的最近点。
+        d1 = vector_dot(n1, first.origin_mm)
+        d2 = vector_dot(n2, second.origin_mm)
+        point = vector_scale(vector_add(vector_scale(vector_cross(n2, direction), d1), vector_scale(vector_cross(direction, n1), d2)), 1 / squared_length)
+        # 6. 由该点与归一化的交线方向组合成基准轴 AxisSpec。
+        axis = AxisSpec(origin_mm=point, direction=vector_unit(direction, field_name="reference axis"))
+    # 7. 注册为拓扑上下文，并返回结果对象（携带该轴）。
+    session.topology.register_context(node.feature_id, axis)
+    return session.result(node, context=axis)