improve `globalRecommendedHPABehavior` to be similar to the default one

[sanposhiho]

https://github.com/mercari/tortoise/blob/main/pkg/hpa/service.go#L279-L298

var globalRecommendedHPABehavior = &v2.HorizontalPodAutoscalerBehavior{
	ScaleUp: &v2.HPAScalingRules{
		Policies: []v2.HPAScalingPolicy{
			{
				Type:          v2.PercentScalingPolicy,
				Value:         100,
				PeriodSeconds: 60,
			},
		},
	},
	ScaleDown: &v2.HPAScalingRules{
		Policies: []v2.HPAScalingPolicy{
			{
				Type:          v2.PercentScalingPolicy,
				Value:         2,
				PeriodSeconds: 90,
			},
		},
	},
}

---

Change the above current behaviour to the following, which is more similar to the default
https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/#default-behavior

var globalRecommendedHPABehavior = &v2.HorizontalPodAutoscalerBehavior{
	ScaleUp: &v2.HPAScalingRules{
		Policies: []v2.HPAScalingPolicy{
			{
				Type:          v2.PercentScalingPolicy,
				Value:         100,
				PeriodSeconds: 60,
			},
			{ // added.
				Type:          v2.PodsScalingPolicy,
				Value:         4,
				PeriodSeconds: 60,
			},
		},
	},
	ScaleDown: &v2.HPAScalingRules{
		Policies: []v2.HPAScalingPolicy{
			{
				Type:          v2.PercentScalingPolicy,
				Value:         2,
				PeriodSeconds: 90,
			},
		},
	},
}

[sanposhiho]

If you just change the implementation only, the integration tests would fail.
https://github.com/mercari/tortoise/blob/main/controllers/tortoise_controller_test.go

You have to understand our e2e tests and eventually modify the test cases.
See https://github.com/mercari/tortoise/blob/main/docs/contributor-guide.md#e2e-tests.

[shreeharshshinde]

Thanks for the detailed explanation on this issue.

I initially overlooked how this change impacts the controller integration tests. After going through the controller tests and updating the HPA behavior accordingly, I’ve opened a PR that aligns the default scale-up behavior with Kubernetes

[shreeharshshinde]

While I had one follow-up question ,

beyond matching the Kubernetes default behavior, is there room for Tortoise in the future to optimize scale-up behavior based on workload characteristics (for example, pod startup time or historical traffic spikes),

or is the general goal to stay aligned with Kubernetes semantics unless there is a strong reason to diverge?