Add cleaning mechanism for runners-cache-X machines

We've received today an alert:

Really low disk space left on /opt/gitlab on runners-cache-5.gitlab.com: 1.293%

after checking the machine it's indeed a 1% left, but this 1% is really big 😉

root@runners-cache-5:/opt/gitlab# df -h /opt/gitlab
Filesystem      Size  Used Avail Use% Mounted on
/dev/md0        3.9T  3.6T   51G  99% /opt/gitlab

The usage is quite high:

It's ~667GB in last 30 days! Looking on what uses the space:

root@runners-cache-5:/opt/gitlab# du -hsc *
3.6T    cache
81G     registry
3.6T    total

So the most of available space is used by cache.

I remember we had in the past a mechanism that was clearing the old data, but I can't find any. We should prepare a new one immediately. For cache it should be quite easy - remove files that were not used for more than N days (14?).

But I'm not quite sure how to handle registry data and if we should think about this while it's using only <100GB from a 3.9TB filesystem. @ayufan any suggestions?