[RLlib] EnvRunners fix min/max-metrics window sizes (from inf to config.metrics_num_episodes_for_smoothing).

[simonsays1980]

Why are these changes needed?

The user defines the window size for metrics in metrics_num_episodes_for_smoothing. This needs to be applied to all episode metrics to keep consistency. This PR adds the missing ones for the min/max which would otherwise report something different than the means.

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 :(

[sven1977]

Awesome! That's why I always felt that min/max stats were lifetime (rather than win=100) :D
Makes perfect sense. Thanks for this fix @simonsays1980 !

[sven1977]

LGTM. Thanks a ton for this important fix @simonsays1980

[sven1977]

Ah, wait, one fix still necessary. We need to do the same for MultiAgentEnvRunner. Could you change it there are well?

        # For some metrics, log min/max as well.
        self.metrics.log_dict(
            {
                EPISODE_LEN_MIN: length,
                EPISODE_RETURN_MIN: ret,
            },
            reduce="min",
            window=self.config.metrics_num_episodes_for_smoothing,
        )
        self.metrics.log_dict(
            {
                EPISODE_LEN_MAX: length,
                EPISODE_RETURN_MAX: ret,
            },
            reduce="max",
            window=self.config.metrics_num_episodes_for_smoothing,
        )