FLPATH-3558 Automate acm tests

[testetson22]

Integrate the ACM Cluster Service Provider into the DCM E2E stack with automated API tests, compose overlay, and optional ACM/MCE deployment. Refactor deploy-dcm.sh with a declarative provider registry (providers/*.conf) so adding new service providers requires no changes to flag parsing, usage, or compose wiring.

Key Changes

1. Provider registry: Declarative providers/*.conf files replace hardcoded per-provider logic in deploy-dcm.sh. Flags, usage, compose args, validation, and env exports are all generated from the registry. Current providers: kubevirt, k8s-container, acm-cluster
2. ACM Cluster SP E2E tests: Health endpoint (schema, status, uptime), registration with SPRM, input validation (empty body, missing fields, wrong types, RFC 7807), CRUD basics (list, 404 get/delete) — gated by HyperShift CRD availability
3. ACM/MCE deployment: --deploy-acm / --deploy-mce install ACM or MCE on the OCP cluster. Idempotent — detects existing installations, creates missing CRs, or runs the full upstream script as needed. Clones acm-cluster-service-provider repo for hack/deploy-acm-mce.sh
4. ACM compose overlay: tests/compose-acm-cluster-sp.yaml adds the ACM SP container with kubeconfig mount, pull secret, and NATS wiring. Auto-injected by the provider registry
5. Deploy script enhancements: ensure_podman_running() auto-starts the Podman machine; validate_acm_cluster_provider() creates the target namespace and resolves the pull secret from the cluster's global openshift-config/pull-secret
6. Test harness integration: run-e2e.sh forwards --acm-cluster-service-provider, --deploy-acm/mce, and --acm-cluster-sp-repo/branch flags; exports DCM_ACM_CLUSTER_SP_URL
7. CI: script validation workflow: .github/workflows/validate-scripts.yaml validates flag parsing, provider registry loading, mutual exclusivity (--deploy-acm + --deploy-mce), required-arg enforcement, and unknown flag rejection for both deploy-dcm.sh (11 cases) and run-e2e.sh (5 cases) — runs without deployment infrastructure

---

ACM Cluster SP Test Coverage

Per upstream spec PR #2:

Status	Tests
Automated	Registration with SPRM (cluster provider), health endpoint (schema, status, uptime progression), input validation (empty body, missing fields, wrong types, RFC 7807), CRUD basics (list, 404 get, 404 delete)
Gated	CRUD lifecycle tests require HyperShift CRDs — skip gracefully without ACM
Follow-up	NATS status events — SP monitoring not yet active in container deployment; upstream version matrix (DefaultCompatibilityMatrix) needs OCP 4.19+ entries

---

Provider Registry

Service providers are defined declaratively in providers/*.conf:

Provider	Config	Compose	Validation
KubeVirt	providers/kubevirt.conf	Profile in api-gateway compose.yaml	CNV CRDs + namespace
K8s Container	providers/k8s-container.conf	Profile + tests/compose-sp-test.yaml overlay	Namespace creation
ACM Cluster	providers/acm-cluster.conf	tests/compose-acm-cluster-sp.yaml overlay	Namespace + pull secret

Adding a new provider requires only a .conf file and (optionally) a validation hook function in deploy-dcm.sh.

---

ACM/MCE Deployment Flow

--deploy-acm / --deploy-mce follows this idempotent flow:

1. Check if the CR (MultiClusterHub or MultiClusterEngine) is already Running → skip
2. Check if the operator CSV is installed but CR is missing → create CR
3. Check if the operator CSV is installed and CR exists but not Running → wait
4. Nothing installed → clone upstream repo, run hack/deploy-acm-mce.sh
5. Wait for CR to reach Running (default timeout: 1200s via DEPLOY_TIMEOUT)

---

No redundancy with upstream/product repo tests. The ACM Cluster SP repo (dcm-project/acm-cluster-service-provider) has extensive unit and integration tests (health, registration, handler, monitoring, cluster operations, config, status mappers) but these all test internal Go packages in isolation or with mocked dependencies. Our tests are external E2E tests that hit the running SP over HTTP through the compose stack -- a fundamentally different test layer.

Area	Upstream (unit/integration)	Our E2E tests
Health	health_unit_test.go, health_integration_test.go	Health endpoint schema, status values, uptime progression over real HTTP
Registration	registration_unit_test.go, registration_integration_test.go	Verifies the SP actually registers as service_type=cluster via the gateway's /providers API
Input validation	handler_unit_test.go, server_unit_test.go	POST /clusters with empty body, missing fields, wrong types, RFC 7807 response -- over the network
CRUD	server_integration_test.go, cluster_test.go	GET/DELETE 404s for non-existent clusters, list endpoint -- gated by HyperShift CRDs on a real cluster

The upstream tests validate internal correctness; ours validate the deployed service behaves correctly as part of the full DCM stack. No overlap.

Checklist

• Provider registry (providers/*.conf) replaces hardcoded provider logic
• ACM Cluster SP E2E tests (health, registration, validation, CRUD)
• ACM compose overlay with kubeconfig, pull secret, NATS wiring
• ACM/MCE idempotent deployment (--deploy-acm, --deploy-mce)
• ensure_podman_running() auto-starts Podman machine
• validate_acm_cluster_provider() creates namespace + resolves pull secret
• make test-acm-sp target
• Test harness forwards ACM flags and exports DCM_ACM_CLUSTER_SP_URL
• CI workflow validates script flags and provider registry (16 test cases)
• Documentation updated (README, CLAUDE.md, Cursor rules, compose comments)
• ShellCheck clean
• go vet clean