Split static and dynamic contact params

[diegoferigo]

This PR introduces the support to separate contact parameters in two categories:

• The static parameters are stored in the contact model class. A vectorized computation must share the same static parameters.
• The dynamic parameters are stored in the contact parameters class. These parameters are copied in JaxSimModelData and the user logic can change them during runtime (assuming that the corresponding PyTree remains compatible). These are also the types of parameters over which AD can be performed.

This separation is necessary when the computation of contact forces operates over a JIT primitive that requires a static argument (e.g. fixing the number of iterations of an iterative solver by converting a while loop to a for loop).

---

📚 Documentation preview 📚: https://jaxsim--257.org.readthedocs.build//257/

[xela-95]

Thanks @diegoferigo !! LGTM! 🚀

[traversaro]

Just to mention something related to naming: sometimes (especially in modelling community, like Modelica/FMI/Simulink/Julia's ModellingToolkit) paramters that can be changed during the simulation are called "tunable parameters", see:

• https://fmi-standard.org/docs/3.0.1/#_variable_attributes
• https://it.mathworks.com/help/sldrt/ug/tune-block-parameters-and-matlab-variables.html
• https://discourse.julialang.org/t/automatic-differentiation-with-subset-of-parameters-in-modelingtoolkit/85917

This is to avoid ambiguity between the system dynamics state and "dynamics parameters".

[diegoferigo]

Just to mention something related to naming: sometimes (especially in modelling community, like Modelica/FMI/Simulink/Julia's ModellingToolkit) paramters that can be changed during the simulation are called "tunable parameters", see:

* https://fmi-standard.org/docs/3.0.1/#_variable_attributes

* https://it.mathworks.com/help/sldrt/ug/tune-block-parameters-and-matlab-variables.html

* https://discourse.julialang.org/t/automatic-differentiation-with-subset-of-parameters-in-modelingtoolkit/85917

This is to avoid ambiguity between the system dynamics state and "dynamics parameters".

It makes sense, thanks for pointing this out @traversaro. JAX has the concept of static function arguments, and I used the term dynamic as its dual. However, I agree that it may cause confusion. With your suggestion, we can name these dynamic parameters (in JAX terms) as tunable parameters.

[flferretti]

Thanks a lot!