[RLlib] Multi-agent test and support envs that sometimes return empty obs-dicts (no agent requesting an action).

[sven1977]

Signed-off-by: sven1977 svenmika1977@gmail.com

Multi-agent test and support envs that sometimes return empty obs-dicts (no agent requesting an action).

• This has already worked when not using connectors (config.rollouts(enable_connectors=False)), but seems to break when connectors are enabled (hence we are using EpisodeV2, see bug below).
• Fixes a EpisodeV2::has_init_obs() bug for when agent_id==0 (if-block should check is not None).
• Also moves multi-agent pendulum learning test into proper location.

Why are these changes needed?

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 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]

nice test. we will try enable_connectors=True with this test merged.
thanks.