VPA - using different metric for historical CPU query

[njuettner]

Which component are you using?:

Vertical Pod Autoscaler

Is your feature request designed to solve a problem? If so describe the problem this feature should solve.:

Currently VPA recommender uses container_cpu_usage_seconds_total for adjusting CPU on pods but this might be often to late.

Describe the solution you'd like.:

Instead of using container_cpu_usage_seconds_total it would be nice using container_cpu_cfs_throttled_periods_total as a historical metric.

historicalCpuQuery := fmt.Sprintf("rate(container_cpu_usage_seconds_total{%s}[%s])", podSelector, p.config.HistoryResolution)