Fix N+1 SQL regression in exporting issues to CSV

https://gitlab.com/gitlab-org/gitlab/-/merge_requests/50449 refactored
`ExportCsvService` to support both issues and requirements. However, it
introduced an N+1 SQL query when loading labels of the list of issues
that significantly slowed down exporting of many issues.

We bring back the memoization for labels inside the
`Issues::ExportCsvService` to avoid this.

Closes https://gitlab.com/gitlab-org/gitlab/-/issues/321622

app/services/issues/export_csv_service.rb

@@ -5,6 +5,12 @@ module Issues
     include Gitlab::Routing.url_helpers
     include GitlabRoutingHelper
 
+    def initialize(issuables_relation, project)
+      super
+
+      @labels = @issuables.labels_hash
+    end
+
     def email(user)
       Notify.issues_csv_email(user, project, csv_data, csv_builder.status).deliver_now
     end
@@ -41,7 +47,7 @@ module Issues
     end
 
     def issue_labels(issue)
-      issuables.labels_hash[issue.id].sort.join(',').presence
+      @labels[issue.id].sort.join(',').presence
     end
 
     # rubocop: disable CodeReuse/ActiveRecord

changelogs/unreleased/sh-fix-export-csv-service-nplusone.yml

+---
+title: Fix N+1 SQL regression in exporting issues to CSV
+merge_request: 54287
+author:
+type: performance