Library/NM: Always rollback checkpoint on failure

[tyll]

If there is a NetworkManager checkpoint, roll it back on any failure.
Otherwise future invocations of the role fail until the checkpoint timed
out.

closes #154

[tyll]

I tested this manually with:
ansible-playbook -i rhel82-cloud, tests/tests_ethtool_features_initscripts.yml tests/tests_ethtool_features_nm.yml tests/tests_states.yml tests/tests_states.yml

[tyll]

@richm @pcahyna just for statistics, the CI is failing again for infrastructure problems, for example:

fatal: [/cache/Fedora-Cloud-Base-31-1.9.x86_64.qcow2]: FAILED! => {"changed": false, "msg": "Failed to download packages: Curl error (6): Couldn't resolve host name for https://mirrors.fedoraproject.org/metalink?repo=fedora-31&arch=x86_64 [Could not resolve host: mirrors.fedoraproject.org]", "results": []}

[thom311]

lgtm