fix(wechat): defer 45047/45015 delivery and control customer message budget

Persist formal replies when WeChat customer-service quota is exhausted,
claim them once on the next passive response, skip progress notices for
page generation, and suppress failure notices while a formal reply is pending.

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

wechat-mp.test.mjs

@@ -1220,7 +1220,7 @@ test('wechat mp service splits long agent replies into multiple customer message
   assert.equal(combined, longReply);
 });
 
-test('wechat mp service surfaces wechat customer-service errcode and errmsg', async () => {
+test('wechat mp service defers 45047 customer-service delivery instead of throwing', async () => {
   const service = createBoundWechatService({
     userAuth: {
       async getWechatOpenidForUser() {
@@ -1241,10 +1241,62 @@ test('wechat mp service surfaces wechat customer-service errcode and errmsg', as
     },
   });
 
-  await assert.rejects(
-    service.sendTextToUser('user-1', 'hello'),
-    /errcode=45047.*out of response count limit/,
+  const result = await service.sendTextToUser('user-1', 'hello');
+  assert.equal(result.deferred, true);
+  assert.equal(result.errcode, 45047);
+});
+
+test('wechat mp service claims deferred delivery on next passive response once', async () => {
+  const token = 'token';
+  const timestamp = '1710000000';
+  const nonce = 'nonce';
+  const service = createBoundWechatService({
+    config: {
+      progressDelayMs: 0,
+      ackText: '已收到',
+    },
+    userAuth: {
+      async getWechatOpenidForUser() {
+        return 'openid-1';
+      },
+    },
+    wechatFetch: async (url) => {
+      if (String(url).includes('/cgi-bin/stable_token')) {
+        return new Response(JSON.stringify({ access_token: 'access-1', expires_in: 7200 }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      return new Response(JSON.stringify({ errcode: 45047, errmsg: 'out of response count limit rid: test-rid' }), {
+        status: 200,
+        headers: { 'Content-Type': 'application/json' },
+      });
+    },
+  });
+
+  const deferred = await service.sendTextToUser(
+    'user-1',
+    '页面已生成：https://example.com/MindSpace/user-1/public/demo.html',
   );
+  assert.equal(deferred.deferred, true);
+  assert.equal(deferred.errcode, 45047);
+
+  const claimed = await service.handleInboundMessage(inboundXml({ content: '继续' }), {
+    timestamp,
+    nonce,
+    signature: signatureFor(token, timestamp, nonce),
+  });
+  assert.equal(claimed.status, 200);
+  assert.match(String(claimed.body), /demo\.html/);
+  assert.doesNotMatch(String(claimed.body), /已收到/);
+
+  const followUp = await service.handleInboundMessage(inboundXml({ content: '再来一条' }), {
+    timestamp,
+    nonce,
+    signature: signatureFor(token, timestamp, nonce),
+  });
+  assert.doesNotMatch(String(followUp.body), /demo\.html/);
+  assert.ok(followUp.task);
 });
 
 test('wechat mp service strips markdown emphasis around outbound links', async () => {