Projects::JobsController#index seems to have an N+1 query problem for job artifacts

For example on https://gitlab.com/gitlab-org/gitlab-ce/-/jobs we can see the following queries being executed:

1.007ms	SELECT "ci_job_artifacts".* FROM "ci_job_artifacts" WHERE "ci_job_artifacts"."job_id" = 51103140 AND "ci_job_artifacts"."file_type" = 1 LIMIT 1
1.004ms	SELECT "ci_job_artifacts".* FROM "ci_job_artifacts" WHERE "ci_job_artifacts"."job_id" = 51103138 AND "ci_job_artifacts"."file_type" = 1 LIMIT 1
1.003ms	SELECT "ci_job_artifacts".* FROM "ci_job_artifacts" WHERE "ci_job_artifacts"."job_id" = 51103139 AND "ci_job_artifacts"."file_type" = 1 LIMIT 1
0.998ms	SELECT "ci_job_artifacts".* FROM "ci_job_artifacts" WHERE "ci_job_artifacts"."job_id" = 51103135 AND "ci_job_artifacts"."file_type" = 1 LIMIT 1
0.984ms	SELECT "ci_job_artifacts".* FROM "ci_job_artifacts" WHERE "ci_job_artifacts"."job_id" = 51105134 AND "ci_job_artifacts"."file_type" = 1 LIMIT 1
0.966ms	SELECT "ci_job_artifacts".* FROM "ci_job_artifacts" WHERE "ci_job_artifacts"."job_id" = 51103136 AND "ci_job_artifacts"."file_type" = 1 LIMIT 1
0.958ms	SELECT "ci_job_artifacts".* FROM "ci_job_artifacts" WHERE "ci_job_artifacts"."job_id" = 51103127 AND "ci_job_artifacts"."file_type" = 1 LIMIT 1

This suggests we're not eager loading the right data in the right place.