CloudLaunch does not correctly delete partially launched instances

If cloudlaunch gets past the launch instance step, but fails when assigning a floating ip, the status log will show an error. However, if you attempt to archive and delete the instance afterwards, the instance is left behind, although the archival occurs correctly.

We should be able to delete all resources created during the launch process.

No file or test is named. Start by tracing the launch-instance, floating-IP assignment, archival, and deletion paths, then reproduce the partial-launch failure; done means every resource created before the failure is removed while archival still succeeds.

Written by the indexing model from the issue text.