Hotfix: updates of autodetected images in HRs

Instead of returning the error when an image map for a container
name (= alias) could not be found, it should fallthrough and attempt
to just update the workload using the container name. If we do not
do this it blocks the updates of automatically detected images, as
reported in #2399.

cluster/kubernetes/manifests.go

@@ -145,18 +145,18 @@ func (m *manifests) SetWorkloadContainerImage(def []byte, id resource.ID, contai
 	// map for the given container to perform an update based on mapped YAML
 	// dot notation paths. If resolving the map fails (either because there
 	// is no map for the given container, or the mapping does not resolve
-	// in to a valid image ref), it returns the error.
+	// in to a valid image ref), it falls through and attempts to update
+	// using just the container name (as it must origin from an automated
+	// detection).
 	//
 	// NB: we do this here and not in e.g. the `resource` package, to ensure
 	// everything _outside_ this package only knows about Kubernetes native
 	// containers, and not the dot notation YAML paths we invented for custom
 	// Helm value structures.
 	if hr, ok := res.(*kresource.HelmRelease); ok {
-		paths, err := hr.GetContainerImageMap(container)
-		if err != nil {
-			return nil, err
+		if paths, err := hr.GetContainerImageMap(container); err == nil {
+			return updateWorkloadImagePaths(def, id, paths, image)
 		}
-		return updateWorkloadImagePaths(def, id, paths, image)
 	}
 	return updateWorkloadContainer(def, id, container, image)
 }