fix(xhs,bilibili): 修复访问受限异常击穿与评论采集边界问题

xhs: PR #958 把 IPBlockError / PlatformAccessError 加入 request() 的
retry_if_not_exception_type 后，tenacity 会直接重抛原异常而不再包装成
RetryError，core 层的 except 分支接不住，单条笔记被限流会让整批
asyncio.gather 抛出，同批已抓取但未入库的数据全部丢失。

- get_note_detail_async_task / get_creators_and_notes 捕获访问受限异常，
  记录明确日志后跳过当前条目，恢复原有的"跳过并继续"语义
- 移除 request() 中已不可达的 IP_ERROR_CODE 分支

bilibili: 修复 get_video_all_comments 的两处翻页边界问题

- is_first_page 改为独立标志，接口返回 next=0 且 is_end=False 时
  不再把后续页误判为首页而重复注入置顶评论
- result 无条件累加，否则开启楼中楼抓取时循环守卫永不推进，
  max_count 完全失效并可能死循环；截断提前到抓取楼中楼之前，
  避免为已被丢弃的评论抓子评论

tests/test_bilibili_client_comments.py

@@ -73,3 +73,79 @@ async def test_video_comments_accept_direct_top_comment():
     )
 
     assert [comment["rpid"] for comment in callbacks] == [1]
+
+
+@pytest.mark.asyncio
+async def test_pinned_comment_not_reinjected_when_cursor_next_stays_zero():
+    """cursor.next 仍为 0 时继续翻页,置顶评论不能被重复注入。"""
+    client = object.__new__(BilibiliClient)
+    collected = []
+    page_calls = 0
+
+    async def get_video_comments(video_id, order_mode, next_page):
+        nonlocal page_calls
+        page_calls += 1
+        # 异常/兜底场景:接口一直回 next=0 且 is_end=False
+        return {
+            "cursor": {"is_end": False, "next": 0},
+            "replies": [{"rpid": 2, "rcount": 0}],
+            "top": {"upper": {"rpid": 1, "rcount": 0}},
+        }
+
+    async def callback(video_id, comments):
+        collected.extend(comment["rpid"] for comment in comments)
+
+    client.get_video_comments = get_video_comments
+
+    await client.get_video_all_comments(
+        video_id="323173868", crawl_interval=0, callback=callback, max_count=5
+    )
+
+    assert page_calls > 1, "该用例需要真正翻过页才有意义"
+    assert collected.count(1) == 1, f"置顶评论被重复采集: {collected}"
+
+
+@pytest.mark.asyncio
+async def test_max_count_applies_when_fetching_sub_comments():
+    """开启楼中楼抓取时 max_count 依然生效,且不为被截断的评论抓子评论。"""
+    client = object.__new__(BilibiliClient)
+    collected = []
+    sub_comment_calls = []
+    page_calls = 0
+
+    async def get_video_comments(video_id, order_mode, next_page):
+        nonlocal page_calls
+        page_calls += 1
+        base = page_calls * 10
+        return {
+            "cursor": {"is_end": False, "next": page_calls},
+            "replies": [
+                {"rpid": base + 1, "rcount": 1},
+                {"rpid": base + 2, "rcount": 1},
+            ],
+        }
+
+    async def get_video_all_level_two_comments(
+        video_id, comment_id, order_mode, ps, crawl_interval, callback
+    ):
+        sub_comment_calls.append(comment_id)
+
+    async def callback(video_id, comments):
+        collected.extend(comment["rpid"] for comment in comments)
+
+    client.get_video_comments = get_video_comments
+    client.get_video_all_level_two_comments = get_video_all_level_two_comments
+
+    result = await client.get_video_all_comments(
+        video_id="323173868",
+        crawl_interval=0,
+        is_fetch_sub_comments=True,
+        callback=callback,
+        max_count=3,
+    )
+
+    # is_end 永远是 False,只能靠 max_count 收敛
+    assert len(result) == 3
+    assert collected == [11, 12, 21]
+    # 第 2 页的 rpid=22 被截断,不应该再去抓它的楼中楼
+    assert sub_comment_calls == [11, 12, 21]