Add slurmgcp-managed infix to resource policy name

[mr0re1]

Motivation: improve collision avoidance, current matching pattern "${cluster_name}-*" is too relaxed.

Consideration for max length of resource policy name, that

Must be a match of regex '(?:[a-z](?:[-a-z0-9]{0,61}[a-z0-9])?)'

• 'slurm_cluster_name' must be a match of regex '^[a-z](?:[a-z0-9]{0,9})$' = 10
• nodeset_name = substr(replace(var.name, "/[^a-z0-9]/", ""), 0, 14) = 14
• assume len(job_id) = 10
• assume len(nodes_chunk_id) = 3
• len("slurmgcp-managed") = 16

len("${cluster_name}-slurmgcp-managed-${nodeset_name}-${job_id}-${nodes_chunk_id}") = 10+1+16+1+14+1+10+3 = 56

Testing:

$ gcloud compute resource-policies list | grep tt0
tt0-slurmgcp-managed-debugnodeset-1-0

$ ./ghpc destroy tt0 $AA
...
module.slurm_controller.null_resource.cleanup_compute[0] (local-exec): Deleted [https://www.googleapis.com/compute/beta/projects/io-playground/zones/us-central1-a/instances/tt0-debugnodeset-0].
module.slurm_controller.null_resource.cleanup_compute[0] (local-exec): Deleted [https://www.googleapis.com/compute/beta/projects/io-playground/zones/us-central1-a/instances/tt0-debugnodeset-1].
module.slurm_controller.null_resource.cleanup_compute[0]: Still destroying... [id=8428718126253180537, 2m20s elapsed]
module.slurm_controller.null_resource.cleanup_compute[0] (local-exec): Deleting resource policies
module.slurm_controller.null_resource.cleanup_compute[0] (local-exec): Deleted [https://www.googleapis.com/compute/beta/projects/io-playground/regions/us-central1/resourcePolicies/tt0-slurmgcp-managed-debugnodeset-1-0].
module.slurm_controller.null_resource.cleanup_compute[0]: Destruction complete after 2m25s
...
$ gcloud compute resource-policies list | grep tt0
$

Followups:

• Include nodeset_name into match pattern of cleanup_compute.sh (once cleanup-per-nodeset is implemented).

Alternative considerations:

• Use existing resource.random_uuid.cluster_id - too long (36 chars); can create unexpected lifecycle dependencies;
• Use dedicated resource.random_uuid.short_cluster_id - can create unexpected lifecycle dependencies;

[tpdownes]

LGTM. Notes:

• 2 of the failing checks appear to be resource name collisions due to recent builds that failed to cleanup
• 2 of the other failing checks are resolved by Delete at most one resource policy at a time #2895 and Hotfix cc version bump #2903 (gcloud does not delete multiple resource policies in a single invocation)