fix(models): 灵犀展示权重不再与「其余渠道」撞号

_PROVIDER_RANK 给 lingxi 7 时 _UNRANKED_PROVIDER 也是 7，未知渠道与
灵犀同级 → 只靠归一键字典序决定先后。把「其余」抬到 8：灵犀恒定排在
raccoon 之后、zen/kilo 等未登记渠道之前。

Author: RobbsLuo

src/api/model_merge.py

@@ -83,7 +83,7 @@
 _PROVIDER_RANK = {"codebuddy": 0, "codebuddy_intl": 1, "trae": 2, "qoder": 3,
                   "codearts": 4, "officeace": 5, "raccoon": 6,
                   "lingxi": 7}
-_UNRANKED_PROVIDER = 7
+_UNRANKED_PROVIDER = 8
 
 # 哨兵名：各上游自己的「自动路由 / 默认模型」占位，语义只在本渠道内成立，不跨
 # 渠道合并。kilo 的 `kilo-auto/free`（上游名 `Auto Free`）与 Qoder 的 `auto`

tests/api/test_models_rank.py

@@ -67,15 +67,18 @@ def import_credential(self, raw):  # pragma: no cover - 未使用
         # 故意让 zen 的模型名字典序最小、CB 的最大，证明排序按渠道权重而非名字
         "codebuddy": Stub("codebuddy", ["z-cb", "a-cb"]),
         "trae": Stub("trae", ["m-trae"]),
+        "lingxi": Stub("lingxi", ["aaa-lingxi"]),
         "zen": Stub("zen", ["aaa-zen"]),
     })
-    for provider_id in ("codebuddy", "trae", "zen"):
+    for provider_id in ("codebuddy", "trae", "lingxi", "zen"):
         app.state.credentials.add(provider=provider_id, credential_data={"t": "x"})
     key = app.state.api_keys.create("root")["api_key"]
     with TestClient(app) as client:
         ids = [item["id"] for item in client.get(
             "/v1/models", headers={"Authorization": f"Bearer {key}"}).json()["data"]]
-    assert ids == ["a-cb", "z-cb", "m-trae", "aaa-zen"]
+    # lingxi 有专属权重（7）→ 排在 zen/kilo 这类「其余」渠道（8）之前，
+    # 即使名字字典序更大
+    assert ids == ["a-cb", "z-cb", "m-trae", "aaa-lingxi", "aaa-zen"]
 
 def test_models_list_orders_dual_provider_by_highest_rank(tmp_path):
     """多渠道模型按最高优先级渠道归位：含 CB 进第一段，仅 TR 的进第二段。"""