Permission steps during updates against an existing instance cause major playbook execution slowdowns & high CPU usage on NFS backed storage

[thunderysteak]

When using the playbook against an existing instance (as part of updates) that is either a part of a larger relay, or has a lot of media files, the file permission tasks crawls media folders as well:
https://github.com/mastodon/mastodon-ansible/blob/main/bare/roles/web/tasks/mastodon-postflight.yml#L102

As Ansible crawls every single media file, it causes the execution to take way too long to perform as it edits every single media file permissions. In testing, it took over two hours and it appeared to be stuck.

This gets worse on NFS backed storage, as Ansible rapidly hammers it and causes the CPU to majorly spike up. Obviously this isn't really ideal.

Mastodon with NFS backed storage Playbook for error replication:
https://github.com/thunderysteak/steaks-ansible-playbooks/tree/main/nfs-mastodon-media

Suggestions on how to solve this:

• Omit /home/mastodon/live/public/system from having chown tasks performed against it
• Same as above but put it behind a feature flag for backwards compatibility with the default behavior being DISABLED.

Tasks that cause issues:

• Ensure that we have correct file permissions with owner being the user and NGINX being the group
• Ensure that we have correct file permissions with owner being the user and www-data being the group
• Ensure that we have correct selinux context permissions for RedHat systems
• Restore selinux condition on RedHat systems
• Create folder structure for Mastodon public folder