ADX-949 revert how foreign keys are handled

[ChasNelson1990]

Description

As detailed in fjelltopp/ckanext-validation#103 and related PRs I fundamentally changed how foreign keys are denoted - at the time this felt like the only solution that was a) general, and so could be pushed upstream, and b) worked with only a minor change to our current approach. However, yesterday it suddenly dawned on me that this actually completely broke our current approach.

The attached is a patch that reimplements some logic from our old fork to solve this problem.

relates fjelltopp/ckanext-validation#105

Checklist

• The Jira ticket for this issue has been updated to "Ready to Review" or equivalent.
• I have developed these changes in discussion with the appropriate project manager.
• My code follows the general Fjelltopp documentation (see Confluence).
• I have made corresponding changes to the Fjelltopp documentation (see Confluence).
• I have rebased this branch with master.
• New dependency changes have been committed.
• I have added automated tests that prove my fix is effective or that my feature works.
• New and existing tests pass locally with my changes.
• My changes generate no new warnings.
• I have performed a self-review of my own code.
• I have assigned at least one reviewer.

[ChasNelson1990]

@tomeksabala - this indirectly impacts the 2024 country estimates (the package schema is pointing to files that have been updated in this PR)