[RLlib; Offline RL] Offline RL example that shows how to customize an offline data pipeline.

[simonsays1980]

Why are these changes needed?

RLlib's new Offline RL was built to be more powerful as its old stack one and competitive to the SOTA standards out in the market. Part of it is its highly customizable pipeline that can be used to literally read any data from
local or remote storage and transform data from raw format to SingleAgentEpisode instances to feed the learner connector pipeline and further to MultiAgentBatch format ready-to-train on the Learner.

This example is the first of multiple ones to show users the cacabilities of RLlib's new Offline RL API. It shows

• how to read in raw image data from a cloud bucket
• how to transform this data via byte streams to numpy arrays
• how to further transform these arrays into SingleAgentEpisode format to be readable by ConnectorV2 pieces
• how to convert further into MultiAgentBatch format to pass to a Learner's update method.

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]

Looks great! Thanks for the quick turnaround @simonsays1980 !

Just the one question on the commented-out line. ??