[serve] Detect node updates

[edoakes]

Why are these changes needed?

Adds an initial control loop that runs every 1s to detect new nodes in the cluster and add routers to them. Also detects removed nodes and removes corresponding actors from them.

Related issue number

Checks

• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/latest/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failure rates at https://ray-travis-tracker.herokuapp.com/.
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested (please justify below)

[AmplabJenkins]

Can one of the admins verify this patch?

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/29190/
Test FAILed.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/29193/
Test FAILed.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/29194/
Test FAILed.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/29192/
Test FAILed.

[simon-mo]

LGTM. Two items we need to do for future work:

1. We should begin to add typing to improve readability. For example, self.routers can be typed as Dict[str, ActorHandle]
2. We should starting to abstract out the checkpoint->start_if_needed->stop_if_needed->checkpoint procedure as it's started to get used in places.

[simon-mo]

Lint is failing on Tune code. I think those are fixed by a471214. Can you merge master?

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/29322/
Test FAILed.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/29378/
Test FAILed.