Explicitly prefer BGS POA on cached appeals attributes (#13554)

connects #13384 

### Description
We prefer displaying BGS POA name, so explicitly cache that. Otherwise, the `PowerOfAttorney` model will trigger loads of VACOLS `case_record` and `representative` that we will not use.

If BGS POA is not found, fall back to VACOLS.

In my manual tests in production, this greatly reduces calls to VACOLS, and instead shifts POA info to the BGS service (which is now cached for 30 days in Redis).

app/jobs/update_cached_appeals_attributes_job.rb

@@ -7,6 +7,7 @@ class UpdateCachedAppealsAttributesJob < CaseflowJob
   include ActionView::Helpers::DateHelper
 
   queue_with_priority :low_priority
+  application_attr :queue
 
   APP_NAME = "caseflow_job"
   METRIC_GROUP_NAME = UpdateCachedAppealsAttributesJob.name.underscore
@@ -101,14 +102,16 @@ def cache_legacy_appeals
   def cache_legacy_appeal_postgres_data(legacy_appeals)
     values_to_cache = legacy_appeals.map do |appeal|
       regional_office = RegionalOffice::CITIES[appeal.closest_regional_office]
+      # bypass PowerOfAttorney model completely and always prefer BGS cache
+      bgs_poa = BgsPowerOfAttorney.new(file_number: appeal.veteran_file_number)
       {
         vacols_id: appeal.vacols_id,
         appeal_id: appeal.id,
         appeal_type: LegacyAppeal.name,
         closest_regional_office_city: regional_office ? regional_office[:city] : COPY::UNKNOWN_REGIONAL_OFFICE,
         closest_regional_office_key: regional_office ? appeal.closest_regional_office : COPY::UNKNOWN_REGIONAL_OFFICE,
         docket_type: appeal.docket_name, # "legacy"
-        power_of_attorney_name: appeal.representative_name,
+        power_of_attorney_name: bgs_poa.representative_name || appeal.representative_name,
         suggested_hearing_location: appeal.suggested_hearing_location&.formatted_location
       }
     end

app/models/power_of_attorney.rb

@@ -10,9 +10,7 @@
 # and lets the user modify VACOLS with BGS information
 # (but not the other way around).
 #
-# TODO: we query VACOLS when the vacols methods are
-# called, even if we've also queried VACOLS outside of this
-# model but in the same request. is this something we should optimize?
+
 class PowerOfAttorney
   include ActiveModel::Model
   include AssociatedVacolsModel