fix(dashboard): 抽取 report-json 前剥离 <think> 思考过程做容错

推理模型(DeepSeek-R1/QwQ 等)输出的 <think>...</think> 思考里常含
```json/裸 {…}，会被误当候选导致抽到错误/残缺数据。抽取前先剥离 think：
- 前端 report-template.ts::extractReportData 开头 stripReasoning(复用 reasoning.ts，
  兼容完整/多段/流式未闭合/只吐闭合四形态)
- 后端 report_json.py 新增 _strip_reasoning 在 extract_report_json 开头调用
- report-template.test.ts +3 例(均通过)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

frontend-web/src/roundtable-planning/lib/report-template.test.ts

@@ -67,6 +67,32 @@ test("extractReportData：dependsOn 只保留真实存在的席位 id（剔除
   assert.deepEqual(d.seats[0].dependsOn, ["s2"]);
 });
 
+test("extractReportData：剥离 <think> 思考过程后再抽取（思考里的假 JSON 不干扰）", () => {
+  // 思考块里故意混入一个会「像」契约但不该被采用的裸 JSON，正文才是真数据。
+  const withThink = [
+    "<think>",
+    '我先草拟一版：{"title":"草稿不要用","seats":[{"id":"x","name":"废弃"}]}',
+    "再想想……",
+    "</think>",
+    SAMPLE,
+  ].join("\n");
+  const d = extractReportData(withThink)!;
+  assert.equal(d.title, "智能机房选址方案大屏");
+  assert.equal(d.seats.length, 2);
+});
+
+test("extractReportData：只吐闭合 </think>（chat template 注入式）也能抽到正文", () => {
+  const onlyClose = ["模型在此思考成本与风险……", "</think>", SAMPLE].join("\n");
+  const d = extractReportData(onlyClose)!;
+  assert.equal(d.title, "智能机房选址方案大屏");
+});
+
+test("extractReportData：流式未闭合 <think> 尾段不被误抽（返回 null）", () => {
+  // 正文还没出，模型还在思考里写了半个 JSON —— 不应抽出数据。
+  const streaming = '正文前导\n<think>我打算输出 {"title":"在想","seats":[{"id":"a","name":"甲"';
+  assert.equal(extractReportData(streaming), null);
+});
+
 test("extractReportData：无 fence / 坏 JSON / 空数据返回 null", () => {
   assert.equal(extractReportData("没有数据块"), null);
   assert.equal(extractReportData("```report-json\n{broken\n```"), null);