cri-o-*: set systemd limits

[runcom]

@giuseppe @mrunalp @rhatdan PTAL

This is need to avoid hitting some limits when running tests and performance tests, for instance:

Oct 26 16:37:00.030: INFO: At 2017-10-26 16:32:06 +0000 UTC - event for example-1-pv6g7: {kubelet ip-172-18-0-172.ec2.internal} Failed: Failed to pull image "centos/mongodb-32-centos7@sha256:f06dfa5a67ac8ecb485c2545a7410b462c0dc1a98459b8e04d3b512f12e77f08": rpc error: code = 2 desc = Error writing blob: Untar error on re-exec cmd: pipe2: too many open files

We would need a new cri-o-centos:latest system container re-built with this patch (when merged) so that we won't hit any limit in the OpenShift CI. Also, it might be nice if we could ship this fix as well: cri-o/cri-o#1078

@giuseppe could you perhaps rebuild with the service file fix? and then we'll ask @lsm5 to rebuild with the needed CRI-O patch.

For the record, these limits are the ones we use in docker as well

Signed-off-by: Antonio Murdaca runcom@redhat.com

[giuseppe]

@rh-atomic-bot r+ 0f77b43

[rh-atomic-bot]

⚡ Test exempted: pull fully rebased and already tested.

[rhatdan]

@runcom nice work