[RLlib] AlgorithmConfigs: Make None a valid value for methods to set properties; Use new NotProvided singleton, instead, to indicate no changes wanted on that property.

[sven1977]

AlgorithmConfigs: Make None a valid value for methods to set properties; Use new NotProvided singleton, instead, to indicate no changes wanted on that property.

Synopsis:

        >>> from ray.rllib.algorithms.algorithm_config import AlgorithmConfig
        >>> config = AlgorithmConfig()
        >>> # Print out the default learning rate.
        >>> print(config.lr)
        ... 0.001
        >>> # Print out the default `preprocessor_pref`.
        >>> print(config.preprocessor_pref)
        ... "deepmind"
        >>> # Will only set the `preprocessor_pref` property (to None) and leave
        >>> # all other properties at their default values.
        >>> config.training(preprocessor_pref=None)
        >>> config.preprocessor_pref is None
        ... True
        >>> # Still the same value (didn't touch it in the call to `.training()`.
        >>> print(config.lr)
        ... 0.001

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 :(

[maxpumperla]

Great, looks very clean to me.