If some ocp cluster ends up in an inconsistent state and we don't have the original jumphost, we use `destroy_cluster.sh -f -i {{ ocp_infra_id }}` to delete it.

This script needs to have the cluster config available in order to check whether all the resources were deleted.

If just the destroy_cluster stage is used, this causes an issue because a fresh jumphost will not have cluster_config.sh rendered. This only happens in the deploy_ocp stage.

When the cluster_config.sh is rendered, it assumes that the ~/.ssh/id_rsa_shiftstack.pub and PULL_SECRET files exist, but these are not required for destroy_cluster.

One way to fix this issue is to render a temperary cluster_config file that does not have these values present.

There should be some documentation provided for using the destroy_cluster playbook with the `-f` option.

Additionally, since destroy_jumphost tries to first destroy the cluster, it is also effected by this bug.

This issue may commonly occur in CI when using the clean-up stages following a failed run.