The implementation does the following.

On insert:

• Upload the image directly to glance from the URL (long running)
• Create and attach a new volume the same size as the root disk
• Rebuild the server with the image, replacing the contents of the root
    disk
• Delete the image

On eject:

• Assume the attached volume has been rewritten with a new image (an ISO
    installer or IPA)
• Detach the volume
• Create an image from the volume (long running)
• Rebuild the server with the new image
• Delete the volume
• Delete the image

The long running operations are performed in a background thread task.
Only one long running operation (insert or eject) can be performed
concurrently for each server. If a long running operation fails, the
only way to feed that back to the user is by re-raising the error during
the next insert/eject request for that server.

The documentation is updated to describe OpenStack driver specifics.
Also the Redfish spec has deprecated accessing VirtualMedia via Managers
so the documentation is updated to refer via Systems.