[HOLD] Use physical targets for scaled parameter objectives

[mickaelbegon]

Closes #848

Summary

• evaluate MINIMIZE_PARAMETER in physical units by applying each parameter scaling
• support both a named parameter and the aggregate key="all"/None cases
• document that standard state and control penalties already expose unscaled variables
• add regression coverage for heterogeneous parameter scaling

Tests

• pytest tests/shard6/test_penalty.py::test_penalty_minimize_parameter_returns_physical_values tests/shard6/test_penalty.py::test_penalty_minimize_state tests/shard6/test_penalty.py::test_penalty_minimize_torque -q (15 passed)

---

This change is 

[EveCharbie]

@mickaelbegon I don't think it is a good idea to scale only for these specific objectives and not sclae for the others, since the behavior will be different. The problem is that we cannot know how to modify the target for other objectives that involve a scaled variable.