Rethink publishing cycle

[lewisjared]

The problem

Currently users publish data before pushing to CI which seems out of order.

Users should ensure tests pass before publishing.

Definition of "done"

• Think about a best practise approach to ensuring notebooks run successfully before publishing
• Document this process
• Update CI to reflect these changes

Additional context

Perhaps the notebook testing should be published on merge by the CI?

[lewisjared]

The current testing framework needs a rethink as well. It doesn't really test anything other than the notebook produces a reproducible result

[lewisjared]

I believe that this is resolved when migrating to new dataset repositories as the publish process doesn't happen at the MR level, instead via version bumps