[RLlib] Fix schedule validation on new API stack (for config settings like lr or entropy_coeff).

[sven1977]

Fix schedule validation on new API stack (for config settings like lr or entropy_coeff).

• The new API stack has no special _schedule settings for the corresponding main settings, such as lr or entropy_coeff.
• Instead, schedules for these settings must be passed into the "main" setting, e.g. config.training(lr=[[0, 0.0001], [1e6, 0.0000001]]).
• The old stack's lr_schedule or entropy_coeff_schedule are not (and have never been) supported on the new API stack.
• Thus far, on the new API stack, however, the config validation would fail to recognize an erroneous config.training(lr=None, lr_schedule=[[]...]) and instead use the default Adam lr (0.001) under the hood.

This PR fixes the validation and throws an error.

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 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 :(

[simonsays1980]

LGTM.