Zathras is ignoring package install failures

[frival]

We've seen this in several cases (the most troublesome being the libxcrypt-compat failure that breaks SPECcpu2017) but this time it was pretty blatant. I'd failed to configure repos on a bare metal server and when Zathras ran on it, we see this:

TASK [install_packages : install yum packages] *********************************
fatal: [perf-arm-17.perf.eng.bos2.dc.redhat.com]: FAILED! => {"changed": false, "failures": ["No package fio available.", "No package gcc available.", "No package gcc-gfortran available.", "No package git available.", "No package java-21-openjdk-headless available.", "No package libnsl available.", "No package libxcrypt-compat available.", "No package make available.", "No package numactl available.", "No package numactl-devel available.", "No package openblas-openmp available.", "No package openmpi available.", "No package openmpi-devel available.", "No package perf available.", "No package perl available.", "No package perl-Math-BigInt.noarch available.", "No package perl-Math-BigRat.noarch available.", "No package perl-Math-Complex.noarch available.", "No package perl-Types-UUID.noarch available.", "No package pip available.", "No package python3-devel available."], "msg": "Failed to install some of the specified packages", "rc": 1, "results": []}

TASK [install_packages : Report Error] *****************************************
changed: [perf-arm-17.perf.eng.bos2.dc.redhat.com]

That means that the package install step failed completely (i.e. not a single package could be installed, so it's less subtle than just "some" package(s) couldn't be installed) yet Zathras happily continued to march on. Any failure in the package install step should be fatal as the system is not configured properly for our runs and we're just wasting the user's time by marching on ahead.