feat(vm): impl corefraction auto

[yaroslavborbat]

Description

Adds vertical CPU autoscaling for virtual machines. Setting spec.cpu.coreFraction: "auto" lets the VM automatically adjust its CPU share to real load: the fraction is raised when the VM is starved and lowered when it's idle, so users no longer have to size CPU by hand or over-provision to be safe.

Recommendations are produced by the stock VPA recommender, while the VM controller applies them in-place — the fraction changes without evicting the pod or restarting the VM. Autoscaling stays within the Burstable QoS range; opting back to a fixed 100% share is an explicit, restart-bearing action.

This is an EE feature, available only when both the VerticalVirtualMachineAutoscaler and HotplugCPUAndMemoryWithInPlaceResize feature gates are enabled.

Why do we need it, and what problem does it solve?

Picking the right coreFraction manually is guesswork: too low starves the VM, too high wastes cluster capacity. Automatic scaling keeps the CPU share matched to actual usage without user intervention and without VM downtime.

What is the expected result?

Enable both feature gates, then set spec.cpu.coreFraction: "auto" on a VM. Its CPU requests track the load — growing under pressure, shrinking when idle — with changes applied in-place. Setting an explicit percentage disables autoscaling for that VM.

Checklist

• The code is covered by unit tests.
• e2e tests passed.
• Documentation updated according to the changes.
• Changes were tested in the Kubernetes cluster manually.

Changelog entries

section: vm
type: feature
summary: "VirtualMachine CPU share can scale automatically with coreFraction: auto."