feat: add Tencent Cloud TokenHub provider with auto model sync

Fetch GET /v1/models on provider/key switch, activate online chat models,
and accept HERMIT_/CST_ key env aliases plus tencent* provider names.

Co-authored-by: Cursor <cursoragent@cursor.com>

tests/test_models.c

@@ -143,7 +143,15 @@ int main(void) {
       "{\"id\":\"vllm-model\",\"max_model_len\":32768,"
       "\"modalities\":{\"input\":[\"text\"]}},"
       "{\"id\":\"plain\"}]}");
-  g_nresp = 6;
+  /* TokenHub-style: status + non-chat ids */
+  g_responses[6] = RESP(
+      "200 OK",
+      "{\"object\":\"list\",\"data\":["
+      "{\"id\":\"hy3\",\"status\":\"online\"},"
+      "{\"id\":\"gone\",\"status\":\"offline\"},"
+      "{\"id\":\"text-embedding-v1\",\"status\":\"online\"},"
+      "{\"id\":\"deepseek-v4-pro\",\"status\":\"pre-offline\"}]}");
+  g_nresp = 7;
   int port = start_server();
   pthread_t th;
   CHECK(pthread_create(&th, NULL, serve, NULL) == 0);
@@ -185,6 +193,11 @@ int main(void) {
   CHECK(meta[2].context == 32768 && meta[2].vision == 0);
   CHECK(meta[3].vision == -1 && meta[3].context == 0 && !meta[3].has_price);
 
+  n = cst_llm_list_models(base, "sk-test", ids, 8, err, sizeof(err));
+  CHECK(n == 2);
+  CHECK(strcmp(ids[0], "hy3") == 0);
+  CHECK(strcmp(ids[1], "deepseek-v4-pro") == 0);
+
   pthread_join(th, NULL);
   test_sock_close(g_listen);