Prevent scaling up of existing cluster with a different spark image

[elmiko]

If a cluster exists, and oshinko is referencing a particular image stream or pull spec, and that cluster does not have image triggers configured, then it would be possible to update the spark image to a different version and subsequently scale up the cluster. The new workers would have a different spark version from the original workers ...

What if anything should be done about this?

migrated from radanalyticsio/oshinko-rest#93