Description of problem:
The pulpcore-api gunicorn process can easily consume up to 3GB+ memory when syncing a large repository (while doing Katello index content).

Steps to Reproduce:
1. Change the gunicorn setting to 1 worker so that we can produce the issue easily.
~~~
Edit "/etc/systemd/system/pulpcore-api.service"

ExecStart=/usr/libexec/pulpcore/gunicorn pulpcore.app.wsgi:application \
--timeout 90 \
-w 1 \ <============== Set this to 1
--access-logfile - \
~~~

2. Restart pulpcore
~~~
systemctl daemon-reload
systemctl restart pulpcore-api
~~~

3. Run the index content against large repository in foreman-rake console
~~~
foreman-rake console
Katello::Repository.find_by_relative_path("Default_Organization/Library/content/dist/rhel/server/7/7Server/x86_64/os").index_content
Katello::Repository.find_by_relative_path("Default_Organization/Library/content/dist/rhel8/8/x86_64/appstream/os").index_content
exit
~~~

Actual results:
~~~

1. After first index content
   pulp XXXXX 21.7 7.7 3021216 2460500 ? S 15:08 2:19 /usr/bin/python3.8 /usr/bin/gunicorn pulpcore.app.wsgi:application --timeout 90 -w 1

1. After second index content
   pulp XXXXX 10.9 11.0 4081824 3521108 ? S 15:08 3:46 /usr/bin/python3.8 /usr/bin/gunicorn pulpcore.app.wsgi:application --timeout 90 -w 1

Expected results:
Less memory consumption