chart: increase controller targetMemoryUtilizationPercentage #14088
Description
Continuing from #10371 (comment)
I suggest we slightly increase the helm chart's controller targetMemoryUtilizationPercentage default to 70%.
At the current 50% HPA default on a 90Mi default request, many newcomer ingress-nginx deployments very quickly or even immediately burst-autoscale to maximum replicas, even during operations unrelated to memory-intensive proxy_buffers. This could give users a wrong impression that this is actually what's normally always supposed to happen on a new deployment of ingress-nginx.
Naturally, it is not the responsibility of kubernetes/ingress-nginx maintainers to manage all newcomer ingress-nginx deployment or even define what the "sensible defaults" deployment should be.
However, I do think that at least a slight target increase would delay any burst-autoscaling on more naive deployments, ensuring that new users avoid erroneously attaining an impression that ingress-nginx is usually just supposed to quickly reach max replicas when deployed through an autoscaling-enabled Helm chart with other defaults left unchanged.