Add retries to native full CI stage

Noticed a few flaky failures in FreeBSD ARM worker. We see those in the docker
workers but there we have retries, so they don't stop the build.

Also, noticed on the s390x worker, we had two flaky failures with eunit crashing
towards the end of the runs and there was not enough time to re-run a 3rd time
as the job died due a timeout, so expand all the timeouts a bit. They are
clearly not enough for all 3 timeouts.