[RLlib; Offline RL] Offline recording new api stack

[simonsays1980]

Why are these changes needed?

The new API stack has no recording possibility, yet, to store experiences to disk. This PR proposes this feature by overriding the EnvRunner to not only sample results but also store them to disk (or cloud storage) with different storing methods. It does so by overriding sampling to

1. Store EpisodeType episodes directly,
2. Or column data generated during sampling parallel to episodes.

Furthermore, it uses ray.data natively, i.e. users can define how the data should be stored by passing in one of ray.data's writing methods (see https://docs.ray.io/en/latest/data/api/input_output.html).

The idea behind overriding the EnvvRunner for sampling is to write datasets in parallel from workers and not sending data around to the driver. This enables to store massive data blazingly fast. This will be needed for offline RL and is specifically handy when running on the cloud writing into blob storage.

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]

LGTM now ... Thanks @simonsays1980 !