The machines profile watcher was missing the ability to tie itself
to the provisioner task life-cycle, which means that when a
provisioner task was killed or restarted then the watcher was still
spinning and never cleaned up, leaking as a resource.

There is another problem that @jameinel spotted in the creation of
the provisioner task, is that if the provisioner errored out on
the construction of the provisioner task, then we never killed the
other watchers and they could also be leaked. We should look at
those in the near future to ensure that any watchers that die on
construction are also cleaned up.

juju#10341

## Description of change

The machines profile watcher was missing the ability to tie itself
to the provisioner task life-cycle, which means that when a
provisioner task was killed or restarted then the watcher was still
spinning and never cleaned up, leaking as a resource.

There is another problem that @jameinel spotted in the creation of
the provisioner task, is that if the provisioner errored out on
the construction of the provisioner task, then we never killed the
other watchers and they could also be leaked. We should look at
those in the near future to ensure that any watchers that die on
construction are also cleaned up.

## QA steps

Ensure that the following test passes:

```console
go test -v ./worker/provisioner/... -check.v -check.f=ProvisionerTaskSuite.TestStartStop
```

## Bug reference

https://bugs.launchpad.net/juju/+bug/1833155

… not satisfy errors.IsNotFound.

…e missing it.

…e missing it.

…exhausted.

…rors-25

juju#10354

## Description of change

Occasionally, juju does not corectly check provider 'not found" errors which get propagated and are surfacing to the users causing confusion.

This PR improves error checks in provider code. I have examined all providers and am fairly certain that I have covered all occurrences.
 
## QA steps

1. bootstrap on gce
2. deploy something sizable.. I've used ```juju deploy cs:canonical-kubernetes```
3. once things settle, run ```juju kill-controller --debug -y -t0 <your controller> --show-log``` and watch the logs...

Before this PR, you'd get a few ```ERROR juju.provider.gce.gceapi conn_instance.go:131 while removing instance "juju-71b3b8-7": googleapi: Error 404: The resource 'projects/juju-239814/zones/us-east1-b/instances/juju-71b3b8-7' was not found, notFound```
After this PR, you will not get any ERROR (if killing is successful)

## Bug reference

https://bugs.launchpad.net/juju/+bug/1831527

# Conflicts:
#	worker/provisioner/provisioner_task.go

# Conflicts:
#	worker/provisioner/provisioner_task.go