fix: run post cleanup after failed jobs

[Piotr1215]

Why

The action advertises its post step as exporting vind container logs and deleting the cluster, but post-if: success() skips that step precisely when a later test fails. Failed E2E jobs therefore lose the diagnostic artifact and leave cleanup to runner destruction.

This was found while dogfooding setup-vind for loft-router's Platform-through-router production canary.

What changed

• Run the JavaScript post action under always() so log export and deletion execute after successful, failed, or cancelled jobs.
• Add a regression test that reads the published action metadata and pins this lifecycle contract.
• Clarify the behavior in the README.

TDD evidence

Before the metadata change, the new test failed with:

Expected pattern: post-if: always()
Received: post-if: success()

After the change:

• npm test -- --runInBand: 15 tests passed.
• npm run build: both checked-in action bundles rebuilt cleanly with no diff.
• git diff --check: clean.

npm run lint is currently not runnable on main because the declared script invokes ESLint but ESLint is absent from package.json and the lockfile. That pre-existing packaging issue is intentionally outside this focused lifecycle fix.

Related: loft-sh/loft-router#23