At the moment we are not running any integration test we have for the
kubernetes executor, which is less then ideal. If we have tests running
we could have caught
https://gitlab.com/gitlab-org/gitlab-runner/-/merge_requests/2123 during
development. We only caught this by accident.

This is the first iteration, the next iteration should improve on:

- Add to [code coverage
  count](https://gitlab.com/gitlab-org/gitlab-runner/-/blob/c7438d7f/.gitlab/ci/coverage.gitlab-ci.yml#L8-20)
- Add tests to [junit
  report](https://gitlab.com/gitlab-org/gitlab-runner/-/blob/c7438d7f/.gitlab/ci/coverage.gitlab-ci.yml#L22-33)
- Only run the integration tests with a build tag as proposed in https://gitlab.com/gitlab-org/gitlab-runner/-/issues/25768#note_345944558
- Run the `TestInteractiveTerminal` test inside of CI https://gitlab.com/gitlab-org/gitlab-runner/-/issues/26421

Use a service image used by configure team
https://gitlab.com/gitlab-org/cluster-integration/test-utils/k3s-gitlab-ci
which exposes the k3s config as a http endpoint to make it easy to
connect to via CI. This project is used by them for running [integration
tests](https://gitlab.com/gitlab-org/cluster-integration/cluster-applications/-/blob/28ed97e4a2c7919dc211825160f78d986b33d33e/.gitlab/ci/test.gitlab-ci.yml#L30-42)

reference https://gitlab.com/gitlab-org/gitlab-runner/-/issues/25768