feat: queue_set_waiter_defer — immediate or call_soon callback mode

queue_waiter_handle stores defer_q/cb/arg, no extra malloc.
queue_set_waiter_defer(q, 1) enables call_soon dispatch for waiter callbacks.
bbr_integration uses it instead of manual send_waiter_cb wrapper.

tests/bbr_integration/test_bbr_integration.c

@@ -177,10 +177,9 @@ static void send_timer_cb(void* arg) {
     send_burst((struct test_ctx*)arg);
 }
 
-static void send_waiter_cb(struct ll_queue* q, void* arg) {
+static void send_wtr_cb(struct ll_queue* q, void* arg) {
     (void)q;
-    struct test_ctx* ctx = (struct test_ctx*)arg;
-    uasync_call_soon(ctx->ua, ctx, send_timer_cb);
+    send_burst((struct test_ctx*)arg);
 }
 
 static void send_burst(struct test_ctx* ctx) {
@@ -191,7 +190,7 @@ static void send_burst(struct test_ctx* ctx) {
     if (conn->normalizer && conn->normalizer->input) {
         if (queue_entry_count(conn->normalizer->input) >= SEND_QUEUE_THRESHOLD) {
             queue_waiter_wait(conn->normalizer->input, &ctx->waiter,
-                              send_waiter_cb, ctx);
+                              send_wtr_cb, ctx);
             return;
         }
     }
@@ -391,8 +390,10 @@ int main(void) {
     t0 = now_us();
     while ((now_us() - t0) < 500000ULL) uasync_poll(ctx.ua, 1);
 
-    if (ctx.sender->connections && ctx.sender->connections->normalizer && ctx.sender->connections->normalizer->input)
+    if (ctx.sender->connections && ctx.sender->connections->normalizer && ctx.sender->connections->normalizer->input) {
         queue_set_threshold(ctx.sender->connections->normalizer->input, 0, 0);
+        queue_set_waiter_defer(ctx.sender->connections->normalizer->input, 1);
+    }
 
     /* Start test */
     printf("\n=== Starting traffic (%d seconds) ===\n\n", TEST_DURATION_MS / 1000);