BATS: helm-install-rancher: wait for Available instead of helm --wait

The rancher Deployment uses the default `progressDeadlineSeconds` (10m),
which is shorter than the initial `rancher/rancher` image pull (~487 MB)
on slow networks (notably WSL2 on Windows). When that deadline lapses,
Kubernetes marks the Deployment Failed with `ProgressDeadlineExceeded`;
`helm --wait` reads that condition and bails out before its own
`--timeout` would have allowed the pull to complete.

Drop `helm --wait` and wait on `Available` directly with
`kubectl wait --for=condition=Available --timeout=30m`. `Available` is
driven by Ready replica count and is unaffected by
`ProgressDeadlineExceeded`, so once the pull eventually completes and
the pod goes Ready the wait succeeds.

Also `--set replicas=1` since this is a single-node cluster — running
three Rancher replicas adds no test coverage and consumes scarce VM
memory.

Signed-off-by: Jan Dubois <jan.dubois@suse.com>

Author: jandubois

bats/tests/k8s/helm-install-rancher.bats

@@ -112,9 +112,17 @@ deploy_rancher() {
         --version "$rancher_chart_version" \
         --namespace cattle-system \
         --set hostname="$host" \
-        --wait \
-        --timeout=10m \
+        --set replicas=1 \
         --create-namespace
+
+    # Don't use `helm --wait`: the chart's Deployment relies on the default
+    # `progressDeadlineSeconds` (10m), which is shorter than the initial
+    # rancher image pull on slow networks. Wait on `Available` directly,
+    # since that condition is driven by pod Ready state and is unaffected
+    # by ProgressDeadlineExceeded.
+    kubectl wait --namespace cattle-system \
+        --for=condition=Available --timeout=30m \
+        deployment/rancher
 }
 
 verify_rancher() {