[Core] Fix test_no_worker_child_process_leaks

[pcmoritz]

Why are these changes needed?

Thanks @iycheng for debugging this!

In #33976 we don't always guarantee that children of worker processes are killed -- namely if the worker dies in an unexpected way, they are not (see #26118).

This unfortunately happens in the test_no_worker_child_process_leaks test case since the Ray cluster is torn down when the driver exits, which can lead to worker processes being forcefully terminated [the solution in this PR is to not initialize the Ray cluster in the driver script but start a separate head node with .add_node].

As a result of this forceful killing, the test becomes flaky (it passes if the workers are killed through the normal termination, and it doesn't pass if the workers are killed otherwise). The flakiness is not super big right now but becomes big with some other changes.

This should eventually be fixed with #26118.

Related issue number

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• 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/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[cadedaniel]

nice find and fix! thanks @pcmoritz and @iycheng

[fishbone]

I also created a P1 issue which shouldn't be too hard to fix when worker decide to exit itself.