2.1 don't fallback on error

[jameinel]

Description of change

We've had a number of bugs over time about containers ending up on the wrong bridge. One reason for this is that we have code that if there is an error while setting up a container, it falls back to using 'lxdbr0'. It did this because that was the only way to get containers on places where we were unable to ask the provider for extra IP addresses. We've since fixed the code to be more logical about it, and explicitly request 'lxdbr0' instead of falling back to it.

This removes the 'on error fallback' code, and finds one more edge case where the fallback was triggering. Namely on AWS we'll see lxdbr0 but we may not see an IP address on the bridge. Which was causing us to not create a DHCP allocated address for eth0.

QA steps

This should only really trigger when there is an error, but we should be sure that:

 $ juju bootstrap aws
 $ juju switch controller
 $ juju deploy ubuntu --to lxd:0
 $ juju status
 $ juju show-machine 0

Should show the container come up and get a 10.0.0.* address.

Documentation changes

Not directly user facing.

Bug reference

https://pad.lv/1656326

[jameinel]

$$merge$$

[jujubot]

Status: merge request accepted. Url: http://juju-ci.vapour.ws:8080/job/github-merge-juju