fix: smooth APPO Go1 joystick actions

[Mario-Ancx]

Summary

• add a second-order action smoothness reward helper for locomotion actions
• register Go1 action_smooth and existing contact rewards in the Go1 reward dispatch
• add the APPO Go1 Motrix owner config and enable action_smooth there
• refresh the generated support matrix for APPO Go1 Motrix coverage

Linked Work

• Issue: Fixes Bug: APPO Go1 joystick Motrix action jitter #619
• Milestone: none

Validation

• make check
• uv run pytest tests/config/test_config_system.py::test_supported_task_composes tests/scripts/test_support_matrix.py tests/scripts/test_check_docs.py::test_documentation_files_match_current_repo_contracts -q
• Additional task-specific validation listed below

Commands actually run:

make check
uv run pytest tests/config/test_config_system.py::test_supported_task_composes tests/scripts/test_support_matrix.py tests/scripts/test_check_docs.py::test_documentation_files_match_current_repo_contracts -q
uv run scripts/generate_support_matrix.py --write

Impact

• Backend impact: both
• Platform impact: Linux / macOS expected neutral
• Training effect expected: yes, APPO Go1 Motrix gets an action smoothness penalty to reduce high-frequency action changes

Artifacts

• W&B: none
• benchmark result: none
• video / screenshot: local eval videos were generated for the compared Go1JoystickFlat APPO runs
• ONNX / checkpoint: none committed

Checklist

• Added or updated tests where needed
• Updated docs if behavior or workflow changed
• Linked the driving issue
• Noted any follow-up work explicitly

Follow-up work: compare new Motrix training run videos after retraining with the added action smoothness reward.