Prevent unit failing when starting model migration

[mjs]

When the model migration fortress was locked down some of the manifolds used by the uniter could return a "fortress shutting down" error which was returned to the uniter resolver loop. This would cause the unit to be put into error before the uniter shut down, causing prechecks failures. The problem is somewhat timing dependent which is why it was being seen intermittently.

There are a number of fixes here:

1. The unit agent now logs unclassified resolver loop errors instead of swallowing them.
2. The migrationminion now logs when it is reporting back to the master.
3. The uniter now takes an optional function to support translation of resolver loop errors.
4. The fortress shutdown error is now exported.
5. Using the above fortress errors coming out of the resolver loop now cause the resolver to restart instead of causing a unit error.
6. The migrationmaster now runs prechecks twice in the QUIECSE phase: before minions have reported back to avoid long waits for an agent that will never be able to report and again after minions have reported, to ensure that the model is truly ok to migrate.

QA

Using a charm with a long sleep in its config-changed hook, tested multiple migrations which were triggered while the hook was executing. Previously this would reliably trigger a precheck failure.

Fixes https://bugs.launchpad.net/juju/+bug/1620438

[anastasiamac]

Please forward port to develop (2.2) to when ready.

[mjs]

$$merge$$

[jujubot]

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

[jujubot]

Build failed: Tests failed
build url: http://juju-ci.vapour.ws:8080/job/github-merge-juju/9882

[mjs]

$$mongodb-sucks$$

[jujubot]

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