feat(qq): stream the full turn feed as discrete messages

Show what the agent is actually doing during a turn, the way Claude
Code does: interim explanation text, every Tool call, and the error
detail when a tool fails, each as its own pushed message before the
final markdown answer.

- HarnessReplyTracker.consume now returns every frame of a polling
  batch in order (same-batch text frames collapse to the latest)
  instead of dropping all but the last event, which silently ate
  tool-call frames that shared a batch with their result
- tool frames carry callId; status frames carry the correlated
  toolName and a flattened error text parsed from the tool/result
  error payload (message, or name: code)
- the QQ bridge holds interim step text until the next tool call
  (or the end of the turn) pushes it, emits 'Tool call <name>' per
  call, emits 'Tool call <name>\nError: <reason>' for failures, and
  skips re-sending interim text that already equals the final answer

Other channels keep their existing behaviour: the status frame still
reads '正在整理结果…' and they ignore the new optional fields.

test/channels/feishu/harness-client.test.mjs

@@ -713,7 +713,7 @@ test('HarnessClient delivers an existing file-only Turn directly', async (t) =>
 test('HarnessReplyTracker correlates the prompt and emits only answer text', () => {
   const tracker = new HarnessReplyTracker({ promptRpcId: 'prompt-1', afterSeq: 10 });
 
-  assert.equal(tracker.consume([
+  assert.deepEqual(tracker.consume([
     { event: { type: 'turn/start', seq: 11, data: { turn: 4 } } },
     { event: {
       type: 'user/message',
@@ -726,7 +726,7 @@ test('HarnessReplyTracker correlates the prompt and emits only answer text', ()
       data: { turn: 4, step: 1, chunk: { type: 'text-delta', index: 0, text: '忽略' } },
     } },
     { event: { type: 'turn/end', seq: 14, data: { turn: 4, reason: { kind: 'completed' } } } },
-  ]), null);
+  ]), []);
   assert.equal(tracker.finished, false);
 
   const first = tracker.consume([
@@ -747,7 +747,7 @@ test('HarnessReplyTracker correlates the prompt and emits only answer text', ()
       data: { turn: 5, step: 1, chunk: { type: 'text-delta', index: 1, text: '深圳' } },
     } },
   ]);
-  assert.deepEqual(first, { type: 'text', text: '深圳' });
+  assert.deepEqual(first, [{ type: 'text', text: '深圳' }]);
 
   const second = tracker.consume([
     { event: {
@@ -761,7 +761,7 @@ test('HarnessReplyTracker correlates the prompt and emits only answer text', ()
       data: { turn: 5, step: 1, chunk: { type: 'text-delta', index: 1, text: '明天有雨' } },
     } },
   ]);
-  assert.deepEqual(second, { type: 'text', text: '深圳明天有雨' });
+  assert.deepEqual(second, [{ type: 'text', text: '深圳明天有雨' }]);
 
   const final = tracker.consume([
     { event: {
@@ -778,7 +778,7 @@ test('HarnessReplyTracker correlates the prompt and emits only answer text', ()
     } },
     { event: { type: 'turn/end', seq: 21, data: { turn: 5, reason: { kind: 'completed' } } } },
   ]);
-  assert.deepEqual(final, { type: 'text', text: '深圳明天有阵雨。' });
+  assert.deepEqual(final, [{ type: 'text', text: '深圳明天有阵雨。' }]);
   assert.equal(tracker.finished, true);
   assert.equal(tracker.answer, '深圳明天有阵雨。');
   assert.deepEqual(tracker.reason, { kind: 'completed' });
@@ -791,9 +791,28 @@ test('HarnessReplyTracker emits tool progress without exposing tool results', ()
     { type: 'user/message', seq: 2, data: { source: { rpcId: 'prompt-tool' } } },
     { type: 'tool/call', seq: 3, data: { turn: 1, step: 1, name: 'web_search' } },
   ]);
-  assert.deepEqual(update, { type: 'tool', name: 'web_search' });
+  assert.deepEqual(update, [{ type: 'tool', name: 'web_search' }]);
 
   assert.deepEqual(tracker.consume([
     { type: 'tool/result', seq: 4, data: { turn: 1, step: 1, secret: 'not rendered' } },
-  ]), { type: 'status', text: '正在整理结果…' });
+  ]), [{ type: 'status', text: '正在整理结果…', toolName: 'web_search' }]);
+});
+
+test('HarnessReplyTracker keeps every frame of a batched turn in order', () => {
+  const tracker = new HarnessReplyTracker({ promptRpcId: 'prompt-batch' });
+  const updates = tracker.consume([
+    { type: 'turn/start', seq: 1, data: { turn: 1 } },
+    { type: 'user/message', seq: 2, data: { source: { rpcId: 'prompt-batch' } } },
+    { type: 'assistant/chunk', seq: 3, data: { turn: 1, step: 0, chunk: { type: 'text-delta', index: 0, text: '先创建再观察：' } } },
+    { type: 'tool/call', seq: 4, data: { turn: 1, step: 1, name: 'add_observations' } },
+    { type: 'tool/result', seq: 5, data: { turn: 1, step: 1, error: { message: 'Status code: 404.' } } },
+    { type: 'tool/call', seq: 6, data: { turn: 1, step: 2, name: 'create_entities' } },
+  ]);
+  assert.deepEqual(updates, [
+    { type: 'text', text: '先创建再观察：' },
+    { type: 'tool', name: 'add_observations' },
+    { type: 'status', text: '正在整理结果…', toolName: 'add_observations', error: 'Status code: 404.' },
+    { type: 'tool', name: 'create_entities' },
+  ]);
+  assert.equal(tracker.answer, '先创建再观察：');
 });