chore: 清理违反底座规则的死代码并修正接口文档编号

- 删除生产死代码 app/service/knowledge_tool_service.py 与
  app/infrastructure/milvus_knowledge_adapter.py：后者硬编码 Milvus 字段名，
  违反 AGENTS.md §E，且仅被前者引用；生产检索链路实际走
  knowledge_search_tool -> KnowledgeSearchService -> knowledge_schema 运行时探测。
- 删除上述两模块的单测，以及依赖 legacy 位置参数构造的
  tests/unit/service/test_knowledge_retrieval.py。
- app/service/knowledge_retrieval_service.py 整文件回退底座版本，
  移除 legacy 双构造与重复检索实现。
- docs/05-接口文档.md：客服画像候选改登记为 §8.5，恢复 §8.2 解析知识引用；
  既有 §8.1-§8.4 编号全部保持，修复此前出现两个 8.3 的问题。
- app/model/profile.py：current_customer_id 改为普通可空列映射，与
  alembic/baseline_generated.sql 及真实库一致；原 Computed 声明会让 ORM 把该列
  从 INSERT 中排除，与「必须显式写入」的实际 schema 不符。
- 新增 docs/客服Agent接入底座扩展说明_v1.md，供集成分支评审逐项确认。

验证：pytest tests/unit tests/contract -> 1275 passed, 2 skipped, 0 failed；
ruff check app tests tools alembic 通过；mypy app 通过（244 个源文件）。

app/service/knowledge_retrieval_service.py

@@ -113,25 +113,16 @@ class KnowledgeRetrievalService:
     def __init__(
         self,
         client: Any,
-        *legacy_args: Any,
+        *,
         embedder: Any = None,
         session_factory: Callable[[], Any] | None = None,
         config: KnowledgeRuntimeConfig | None = None,
         vector_dim: int = VECTOR_DIM,
     ) -> None:
-        # 兼容早期客服工具的 positional 构造：embedder, vector_store, config, authority。
-        self._legacy = len(legacy_args) == 3
-        if self._legacy:
-            self.embedder = client
-            self.client = legacy_args[0]
-            self.config = legacy_args[1]
-            self._authority = legacy_args[2]
-        else:
-            self.client = client
-            self.embedder = embedder
-            self.config = config or KnowledgeRuntimeConfig()
-            self._authority = None
+        self.client = client
+        self.embedder = embedder
         self._session_factory: Callable[[], Any] = session_factory or SessionFactory
+        self.config = config or KnowledgeRuntimeConfig()
         self.vector_dim = int(vector_dim)
 
     # --- 入口 -----------------------------------------------------------------
@@ -139,14 +130,11 @@ class KnowledgeRetrievalService:
     async def search(
         self,
         query: KnowledgeQuery,
-        legacy_context: Any = None,
         *,
         embedding_endpoints: Sequence[Any] | None = None,
         embedder: Any = None,
     ) -> KnowledgeSearchResult:
         """执行检索。集合由意图映射，调用方无法指定集合名。"""
-        if self._legacy:
-            return await self._legacy_search(query)
         targets = self._assert_collections_allowed(query.intents)
         top_k = min(int(query.top_k), self.config.result_limit)
         searched = tuple(sorted({collection for collection, _ in targets}))
@@ -162,49 +150,6 @@ class KnowledgeRetrievalService:
         hits = self._to_hits(verified)
         return KnowledgeSearchResult(hits=hits, degraded=False, searched_collections=searched)
 
-    async def _legacy_search(self, query: KnowledgeQuery) -> KnowledgeSearchResult:
-        """兼容旧工具调用，仍复用同一意图路由和 MySQL 权威回查。"""
-        targets = self._assert_collections_allowed(query.intents)
-        collections = tuple(sorted({collection for collection, _ in targets}))
-        top_k = min(int(query.top_k), int(getattr(self.config, "result_limit", 20)))
-        try:
-            raw_vector = await self.embedder.embed(query.query)
-            vector = self._vector_of(raw_vector)
-            self.assert_vector_dim(vector)
-            rows: list[dict[str, Any]] = []
-            for collection, route_top_k in targets:
-                rows.extend(await self.client.search(
-                    collection, vector, min(top_k, route_top_k)
-                ))
-            if self._authority is not None:
-                verified = await self._authority.filter_published(tuple(
-                    KnowledgeHit(
-                        knowledge_id=str(row.get("knowledge_id")),
-                        collection=collection,
-                        title=row.get("title"),
-                        snippet=str(row.get("snippet") or ""),
-                        score=self._score(row),
-                    )
-                    for row in rows
-                    for collection, _ in targets
-                    if str(row.get("collection") or collection) == collection
-                ))
-                return KnowledgeSearchResult(
-                    hits=tuple(verified), searched_collections=collections
-                )
-            return KnowledgeSearchResult(hits=self._to_hits(rows), searched_collections=collections)
-        except Exception:
-            if self._authority is None:
-                return KnowledgeSearchResult(
-                    hits=(), degraded=True, degradation_reason="milvus_unavailable",
-                    searched_collections=collections,
-                )
-            fallback = await self._authority.search_keyword(query, collections, top_k)
-            return KnowledgeSearchResult(
-                hits=tuple(fallback), degraded=True,
-                degradation_reason="milvus_unavailable", searched_collections=collections,
-            )
-
     # --- 路由与白名单 ---------------------------------------------------------
 
     def _assert_collections_allowed(
@@ -257,8 +202,6 @@ class KnowledgeRetrievalService:
 
     @staticmethod
     def _vector_of(execution: Any) -> list[float]:
-        if isinstance(execution, list | tuple):
-            return [float(item) for item in execution]
         raw = getattr(execution, "vector", None)
         if raw is None and isinstance(execution, Mapping):
             raw = execution.get("vector")