Migration to hydra latest

[RedLeader962]

Types of changes

• [x] Docs change / refactoring / dependency upgrade
• [ ] Bug fix (non-breaking change which fixes an issue)
• [ ] New feature (non-breaking change which adds functionality)
• [x] Breaking change (fix or feature that would cause existing functionality to change)

Motivation and Context / Related issue

Motivation: Enable importing mbrl-lib in project requiring newer hydra-core>=1.1.2 version. Context: mbrl-lib hydra-core version was freezed at v1.0.3 which was release in 2020. A lot of useful functionality were introduce since then.

Regression warning/bug that where fixed: Post hydra-core version 1.1 introduce breaking changes that were fixed in this PR the following way:

• hydra-core>=1.1.0 set instantiate fct to recursive by default, breaking all mbrl-lib use of hydra.instantiate original behaviour with nested dict config (hydra 1.1.0 release note). Changing config files on the user side would be error-prone, so I simply added _recursive_=False to every hydra.utils.instantiate(<cfg>) that deal with nested instantiation logic;
• Preserve mbrl-lib default config ordering behaviour by adding _self_ keyword in examples/conf/main.yaml defaults  (Ref https://hydra.cc/docs/1.1/upgrades/1.0_to_1.1/default_composition_order/);
• Removed deprecated package header @package _group_ to all config group yaml file (ref https://hydra.cc/docs/1.1/upgrades/1.0_to_1.1/changes_to_package_header/).

Remark on breaking change:

• overriding via command line now require prefixing with ++ eg: python -m mbrl.examples.main algorithm=mbpo ++device=cpu

Note:

• I've set the minimal requirement to hydra-core>=1.1.2. Older than that wont be compatible with the proposed changes.
• Documentation was updated

How Has This Been Tested (if it applies)

• Repeated all the following test with hydra-core==[1.1.2, 1.2, 1.3.2]
  • Executed pytest tests on both tests/core and tests/algorithms
  • Manually run the modified example at mbrl/examples/README.md

    python -m mbrl.examples.main \
     algorithm=mbpo \
    overrides=mbpo_cartpole \
    dynamics_model=gaussian_mlp_ensemble \
    ++device=cpu \
    ++overrides.sac_batch_size=256 \
    ++overrides.validation_ratio=0.2 \
    ++dynamics_model.activation_fn_cfg._target_=torch.nn.ReLU 

• Execute same command as in .github/workflows/ci.yml
  • Execute dependencies install via requirement file
  • Lint with flake8
  • Lint with mypy
  • Check format with black

Checklist

• [x] The documentation is up-to-date with the changes I made.
• [x] I have read the CONTRIBUTING document and completed the CLA (see CONTRIBUTING).
• [x] All tests passed, and additional code has been covered with new tests.