fix(UME): 兼容首批无marker场景并补充回归测试

告警分页在响应头缺失 marker 时直接结束迭代，避免首批已全量却重复翻页；同时补充对应单测并保持 HTTPTransport 传输方式与现网一致。

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

netx_api/ume_client.py

@@ -177,8 +177,9 @@ class UMEClient:
     def _client(self) -> httpx.Client:
         # Use explicit HTTPTransport to keep behavior consistent with onsite validation.
         # In this mode, requests run over HTTP/1.1 and avoid HTTP/2 negotiation issues.
-        transport = httpx.HTTPTransport()
-        return httpx.Client(verify=self.verify_tls, timeout=self.timeout_s, transport=transport)
+        transport = httpx.HTTPTransport(verify=self.verify_tls, http2=False)
+        return httpx.Client(transport=transport, timeout=self.timeout_s)
+        
 
     def _extract_token_and_ttl(self, payload: dict[str, Any]) -> tuple[str, int | None]:
         token = ""

netx_api/ume_sync_service.py

@@ -282,8 +282,12 @@ def _sync_alarms_common(
             next_marker = str(diag.marker or "").strip()
             if is_end_of_reply:
                 break
-            # marker paging: no marker and empty data means no next page.
-            if not next_marker and not rows:
+            # marker paging: if response header has no marker, treat as end of iteration.
+            # Some UME deployments omit marker when the first page already contains all rows.
+            if not next_marker:
+                if rows:
+                    warnings.append("marker_missing_stop")
+                    paging_note = "marker_missing_stop"
                 break
 
         if not is_uncleared: