I have a cluster of kubernetes, so I want metrics to measure when app need to scale-out. My first idea is to using "Latency" for measure responses from app's user to app's pod ,so I'm not sure Is the kubernetes has something that can represent about user application pods latency by the time or not?
This is not something kubernetes really handles. We don't know what an app is. You might start with your frontend and tracking request-time to response-time. This is imperfect because it excludes a bunch of things that actually matter, like network ingress.
Or you might set up a prober to access your app through the public interface and time that.