On publish, if previous versions are identical, publish fails

[ajkiessl]

We ran into a work: https://scholarsphere.psu.edu/resources/d0e21912-5ca6-4eb4-bc5a-a7c057fb879e that would not publish it's V3 because V1 and V2 are identical. On the attempted publish, it also did not display or record an error. I was only able to determine the issue by pulling the production data down locally and using a debugger to catch the error. The ChangedWorkVersionValidator checks to see if the previous version is identical to a given work version. This is ran as a validation on work versions. I'm assuming on publish, since the work is saved and downstream work versions are saved, the validation runs and fails — even when publishing V3. Also, since the validation is failing for a work version not associated with the publish page, the error isn't presented.

Before we can fix this, we need to determine how we want this to work. I think, ideally, we don't want V1 and V2 to ever end up identical? It looks like some things were edited by an admin in V2 which lead to it being identical to V1. So, we should determine why the validation was bypassed when an admin was editing, and if that's what we want.

Or perhaps we should get the error to display and indicate what needs done to fix it. Or, maybe we shouldn't be validating V1 and V2 when V3 is being published.