Add Molecule CI test for nvidia driver role

Right now this is just a basic "does the install succeed?" test, with no
special verification going on.

- Add auto-generated molecule test
- Configure Molecule to test all of
  - CUDA repo on CentOS 7 and 8
  - CUDA repo on Ubuntu 18.04 and 20.04
  - Canonical repo on Ubuntu 18.04 and 20.04
- Add Github Action to test in CI

tasks/install-redhat.yml

@@ -27,6 +27,7 @@
     environment: "{{proxy_env if proxy_env is defined else {}}}"
   - name: reboot to pick up the new kernel
     reboot:
+    when: not nvidia_driver_skip_reboot
 
 - name: add epel repo gpg key
   rpm_key: