fix(site-import): FetchError 白话化映射（spec §6 行1）+ 预算语义注记

_pipeline 增 FetchError 分支（先于通用异常兜底）：入口页不可达/超时——
最高频真实失败路径——映射白话文案「网站打不开或响应太慢——检查 URL」
+ phase=crawl（不可 retry）；ImportTask docstring 的抓取失败归档口径
同步对齐。crawler 模块 docstring 注记 MAX_PAGES 预算语义（只计成功入编
页，链接农场广度消耗由总超时/并发/频控三层兜底）。

Co-Authored-By: Claude Code <noreply@anthropic.com>

tests/test_site_import_task.py

@@ -433,6 +433,18 @@ class TestTasks:
         with pytest.raises(ValueError):
             await task.start("t1", None, retry_task_id=tid_b)
 
+    async def test_FetchError白话化映射(self, monkeypatch):
+        # spec §6 行1：站点不可达/超时是最高频真实失败路径——不漏进
+        # 通用异常兜底裸抛英文内部信息
+        stubs = _Stubs().install(monkeypatch)
+        stubs.crawl_errors.append(safe_fetch.FetchError(
+            "fetch failed for 'https://a.com/': timed out"))
+        tid = await task.start("t1", "https://a.com/")
+        t = await _wait("t1", tid)
+        assert t.state == "failed"
+        assert t.error == "网站打不开或响应太慢——检查 URL"
+        assert t.failed_phase == "crawl"
+
     async def test_30分钟惰性清理(self, monkeypatch):
         _Stubs().install(monkeypatch)
         tid = await task.start("t1", "https://a.com/")