[Ci-users] CI jobs failing because of Epel metalink issues

Some people reported in #centos-devel that their CI jobs were failing (example https://ci.centos.org/job/minishift-addons-nightly-b2d/49/console) because of yum errors.

As you can see it was happening when using EPEL repository. We have confirmation that EPEL infrastructure had an issue with corrupted repodata everywhere, so it hurted us too. While it was fixed upstream, it can take time for the corrected repodata to be replicated everywhere, and I switched the EPEL mirror we're syncing/pulling from to fix that earlier (the previous mirror we used was still bad)

This should fix your EPEL yum errors for now in CI.centos.org