Refactor: Use page version's element repository rather than a new one

[mamhoff]

The PageVersion model offers us a repository we can use, so let's go
with that. This will also avoid using a scope on a potentially already pre-loaded association.

Checklist

• I have followed Pull Request guidelines
• I have added a detailed description into each commit message

[mamhoff]

Interestingly, the only spot where the element repository from the page version is currently used is when publishing pages. This is done through ActiveJob, and that supports (different to e.g. Sidekiq) passing "objects". Those "objects" however, are still serialized through the GlobalID library, and in the serialization/deserialization process they will lose associated data. The only way to get preloading to work now would be something along these lines in the job:

Alchemy::Page.includes(page_includes).find_by!(id: page.id)

Not pretty, but arguably better to load the page twice (once in GlobalID/ActiveJob, once in our app) than to load all versions, elements, ingredients and so on. What do you think?

[tvdeyen]

Makes total sense. Thanks