Add more details to the invoke call trace.

It is always useful information to know if the TPU invoke is Async or not, and if the GPU invoke on the old path or new path.
This can make it obvious in the perfetto trace.

PiperOrigin-RevId: 517162515

mediapipe/calculators/tensor/inference_calculator_gl_advanced.cc

@@ -141,7 +141,7 @@ InferenceCalculatorGlAdvancedImpl::GpuInferenceRunner::Process(
         }
         // Run inference.
         {
-          MEDIAPIPE_PROFILING(GPU_TASK_INVOKE, cc);
+          MEDIAPIPE_PROFILING(GPU_TASK_INVOKE_ADVANCED, cc);
           return tflite_gpu_runner_->Invoke();
         }
       }));