fix(ci): cleanup cluster resourses for run nightly e2e

[universal-itengineer]

Description

Improve the nested E2E resource preparation step in e2e-matrix so it makes decisions based on the actual cluster state and produces clearer CI logs.

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

Previously, the workflow could overestimate freed resources during VM shutdown and the waiting logic was slow and not always informative. It was difficult to understand from CI logs which VMs were selected for shutdown, what resources they were expected to free, and why the step finally failed. In some cases the workflow also kept waiting when there were no more useful candidates or when all requested VMs were already stopped.

What is the expected result?

The workflow should select only relevant non-stopped VM candidates, print a clear shutdown plan, request shutdowns first, then check the real cluster resource state with shared polling. As a result, CI logs become easier to read, waiting becomes more predictable, and failure cases more clearly show when manual intervention is required.

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: ci
type: fix
summary: cleanup cluster resourses for run nightly e2e