Refactor to straighten out language and improve dfe:analytics:check

[duncanjbrown]

The useful analytics:check task detects when we add and remove fields from applications, but it had a shortcoming: it couldn't tell when a field existed in both the allowlist and the blocklist. This wasn't considered first time around because why would that ever happen. We now know that it does happen, and it happens when an entity wasn't previously in the allowlist but was in the blocklist (because we ignored it), and now we've added it to the allowlist (because we've decided not to ignore it) and now it's appearing in both.

Not necessarily a problem, but it's unprofessional and confusing and it looks bad so it's no longer allowed 🚫 .

This PR also contains some refactoring to the README and methods which refer to the word model. We don't believe in models out here; we believe in entities. Keep usage of model and table_name to an absolute minimum so people don't get confused when we're referring to three somewhat equivalent things by three different names.

🤔 You may notice that whilst this PR makes a point of saying that we send entity events rather than model events, the README still only talks about mixing the Entities behaviour into models. This is because the next PR is going to do away with the manual mixin altogether following very good idea from @misaka that we could do it automatically with information we've got lying around in DfE::Analytics — progress continues on that here. This will fix a bug where we don't send updates for join tables without an associated model.

This PR best reviewed commit by commit.

[misaka]

I found this bit a bit hard to digest. I think if you're going to be refactoring this bit, it would make sense to try to make it easier to comprehend, if possible. Maybe something like this (NB: untested code):

      def self.entity_attributes_from_lists(entity, lists)
        lists.map do |list|
          # for each list of model attrs, look up the attrs for this model
          list.fetch(entity, [])
        end.reduce(:concat)
      end

      def self.entities_with_missing_attributes_from_lists(*lists)
        DfE::Analytics.all_entities_in_application
          .reduce({}) do |entities_missing_attributes, entity|
            attributes_considered = entity_attributes_from_lists(entity, lists)
            model = DfE::Analytics.model_for_entity(entity)

            entities_missing_attributes[entity] = model.attribute_names - attributes_considered
            entities_missing_attributes
          end.reject { |_entity, attributes| attributes.empty? }
      end

      def self.entities_with_surplus_attributes_from_lists(*lists)
        DfE::Analytics.all_entities_in_application
          .reduce({}) do |entities_surplus_attributes, entity|
            attributes_considered = entity_attributes_from_lists(entity, lists)
            model = DfE::Analytics.model_for_entity(entity)

            entities_surplus_attributes[entity] =  attributes_considered - model.attribute_names
            entities_surplus_attributes
          end.reject { |_entity, attributes| attributes.empty? }
      end

[misaka]

Looks good! 🚢

[misaka]

Argh. Wrong circle. 🤦