fix(runner-manager): clean up VMs in error state immediately

[cbartz]

What this PR does

Cleans up runners whose cloud VM is in an ERROR state immediately, instead of waiting for the creation timeout.

_get_platform_runners_to_cleanup previously decided cleanup purely from platform health (online/busy/deletable) plus VM age, never consulting the cloud VMState. An ERROR VM was therefore treated like one still booting and only collected once older than RUNNER_MAXIMUM_CREATION_TIME (~23 min). This adds an explicit rule: any requested runner whose mapped VM has state == VMState.ERROR is eligible for cleanup regardless of age.

Why we need it

An OpenStack VM reported as ERROR never recovers, but it still counted toward the desired runner total, so the slot stayed occupied with a dead VM for ~23 minutes and no replacement was created until cleanup finally evicted it. This leaves a flavor with zero usable capacity for that whole window.

Checklist

• I followed the contributing guide
• I added or updated the documentation (if applicable) — updated docs/changelog.md
• I updated docs/changelog.md with user-relevant changes
• I used AI to assist with preparing this PR
• I added or updated tests as needed (unit and integration) — added a unit case covering immediate cleanup of a recent ERROR-state VM
• If this is a Grafana dashboard: I added a screenshot of the dashboard — n/a
• If this is Terraform: terraform fmt passes and tflint reports no errors — n/a
• If the github-runner-manager application has been changed: The application version number is updated in github-runner-manager/pyproject.toml. — bumped 0.18.1 → 0.18.2

[Copilot]

Pull request overview

This PR updates the runner cleanup decision logic so that platform runners backed by cloud VMs in VMState.ERROR are cleaned up immediately, rather than waiting for the creation timeout, improving recovery time and freeing capacity sooner.

Changes:

• Add an “immediate cleanup” rule for requested runners whose mapped VM is in VMState.ERROR.
• Extend unit test coverage to include an ERROR-state VM cleanup scenario.
• Bump github-runner-manager version and document the user-visible behavior change in the changelog.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
github-runner-manager/src/github_runner_manager/manager/runner_manager.py	Adds ERROR-state VM cleanup rule and updates cleanup selection logic.
github-runner-manager/tests/unit/manager/test_runner_manager.py	Adds a unit test case asserting immediate cleanup for ERROR-state VMs.
github-runner-manager/tests/unit/factories/runner_instance_factory.py	Extends cloud runner factory helper to allow specifying VM state for tests.
github-runner-manager/pyproject.toml	Bumps application version to 0.18.2.
docs/changelog.md	Records the new immediate cleanup behavior for ERROR-state VMs.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

[cbartz]

Force-merged (admin) despite the failing Runner tests across bases (22.04) / test_charm_runner check.

The failure is a pre-existing flaky test unrelated to this PR:

• The failing test is test_otel_collector_endpoint_pre_job_installs_config. It dispatches a workflow, waits for it to complete, then SSHes into the runner — but ephemeral runner VMs are torn down on job completion, so the inspection races the runner-manager cleanup. On 22.04 the cleanup won (get_single_runner found zero VMs → found more than one runners or no runners: []); the identical test passed on the 24.04 base in the same run.
• This PR's code path was never exercised: there were zero VMs in VMState.ERROR across all attempts, so the new errored-VM cleanup logic never ran. The VMs were always ACTIVE and removed via the standard ephemeral-completion cleanup that already exists on main.

Follow-up to de-flake the test: #803.