Multi env seed

[gjoliver]

Why are these changes needed?

Deterministic rollout sampling does not work if a worker runs more than 1 environments.

Related issue number

Closes #17900

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/master/.
• [*] 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 :(

[gjoliver]

I moved the actual seed setting logics out of the constructor into a couple of global util functions to clean things a little bit and reduce the size of the constructor.

I have to update CartPoleWithRemoteParamServer env, because if I read it correctly, this env was hardcoded to be deterministric. So the example test would run deterministically regardless of how we set the seed.

I also have a couple of questions I am not sure about (2 TODOS in the function _update_env_seed). Hope to discuss during the review process.

Thanks for all the helps :)

[sven1977]

Looks great! Thanks for the fix and cleaning up a little :), @gjoliver !

[sven1977]

Nice!