Compare database before and after updates

[grantfitzsimmons]

@melton-jason will elaborate more on this, but to identify potential regressions, we need to have an automated way of comparing a database before and after an update. This would identify several regressions that we missed before, including:

• Duplicates introduced in the schema configuration
• In-place replacement of values without updating timestamps
• Records being erroneously created
• Records being re-inserted into a table as a union (see Redis keys not unique to database when executing 0043_normalize_many_to_many #7964)

We should be able to test this in more complex environments ideally, in an instance with several databases, but at a minimum the counts need to be compared.