worker/docker: Destroy containers from watch routines

[victorges]

What does this pull request do? Explain your changes. (required)
There was a slight race condition on shutdown as the watch routines tried to recreate the containers as we were removing them on shutdown.

Fixing this by keeping the lifecycle management in the watch container routine itself.

Specific updates (required)

• Create a global ctx for the whole DockerManager which is closed on Stop
• Create a sync.WaitGroup that gathers all background watchContainer routines
• Destroy containers from within watchContainer when ctx is closed
• Wait for watchContainer routines to stop before exiting

How did you test each of these updates (required)
The box ™️

Does this pull request close any open issues?
Fixes orphan containers after O shutdown.

Checklist:

• Read the contribution guide
• make runs successfully
• All tests in ./test.sh pass
• README and other documentation updated
• Pending changelog updated

[j0sh]

Seems ok, it's a little confusing with the global context in addition to the per function contexts in Stop / Borrow but I don't really understand this bit enough to offer more constructive feedback, so LGTM

[j0sh]

Should m.ctx.Done() be checked as well or is that implicit in the wait group / done?

[victorges]

Actually the manager's ctx will always be done here, since we call m.stop() in the beginning