[Router] Abort the engine when a client disconnects mid-request (#39461)

Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
Co-authored-by: Kan Wu <wukanustc@gmail.com>

experimental/sgl-router/tests/proxy/chat_routing/reorg.rs

@@ -230,7 +230,19 @@ async fn length_selects_plain_bucket_before_engine_selection() {
     let response = app.clone().oneshot(request(body("hi"))).await.unwrap();
     assert_eq!(response.status(), StatusCode::OK);
     let _ = response.into_body().collect().await.unwrap();
-    assert!(short_worker.captured.lock().unwrap().last_body.is_some());
+    let forwarded: serde_json::Value = serde_json::from_slice(
+        short_worker
+            .captured
+            .lock()
+            .unwrap()
+            .last_body
+            .as_ref()
+            .unwrap(),
+    )
+    .unwrap();
+    assert!(crate::common::is_engine_shaped_rid(
+        forwarded["rid"].as_str().unwrap()
+    ));
     assert!(long_worker.captured.lock().unwrap().last_body.is_none());
 
     let response = app
@@ -293,6 +305,7 @@ async fn pd_picks_both_groups_from_selected_bucket_and_shares_bootstrap() {
     let d: serde_json::Value =
         serde_json::from_slice(decode.captured.lock().unwrap().last_body.as_ref().unwrap())
             .unwrap();
+    assert!(p.get("rid").is_none() && d.get("rid").is_none());
     assert!(p["bootstrap_room"].is_number());
     assert_eq!(p["bootstrap_room"], d["bootstrap_room"]);
     let calls = policy.calls.lock().unwrap();