Persistent Agent doesn't like v1beta1 scheduledworkflow in on-prem scenario

[IronPan]

In a non-gke environment, somehow persistent agent doesn't recognize v1beta1 scheduledworkflow and keep on looking for v1alpha1 scheduledworkflow.

The runs are not propagated back to UI.

E0207 05:39:42.822359 1 reflector.go:205] pkg/mod/k8s.io/client-go@v0.0.0-20180718001006-59698c7d9724/tools/cache/reflector.go:99: Failed to list *v1alpha1.ScheduledWorkflow: the server could not find the requested resource (get scheduledworkflows.kubeflow.org)
E0207 05:39:43.824054 1 reflector.go:205] pkg/mod/k8s.io/client-go@v0.0.0-20180718001006-59698c7d9724/tools/cache/reflector.go:99: Failed to list *v1alpha1.ScheduledWorkflow: the server could not find the requested resource (get scheduledworkflows.kubeflow.org)

We might want to temporarily roll back swf version in here
https://github.com/neuromage/kubeflow/blob/32f6dd72e748cafd7830d24dece0fceeabfa4b96/kubeflow/pipeline/pipeline-scheduledworkflow.libsonnet#L157

[IronPan]

This is caused by a version mismatch. The cluster is deployed using kubeflow HEAD which has v1beta1 scheduledworkflow crd while pipeline persistent agent version is v0.1.8 which looks for v1alpha1 SWF crd.

The issue should be gone after bumping the version to v0.1.9.