Add support for computing the bias acceleration of the center of mass

[diegoferigo]

Similarly to #127, this PR computes the bias linear acceleration of the CoM. It provides the $\dot{J}^l_{\text{CMM}} \boldsymbol{\nu}$ contribution to the CoM linear acceleration that is not dependent on $\dot{\boldsymbol{\nu}}$. In inertial-fixed and mixed representation, we have that $G = G[W]$, while in body-fixed representation, we have that $G = G[B]$.

---

📚 Documentation preview 📚: https://jaxsim--161.org.readthedocs.build//161/

[diegoferigo]

I've narrowed down the failures in CI. The new test is currently failing when:

• The model is floating base.
• The velocity representation is Mixed.
• Then base angular velocity $\boldsymbol{\omega}_{W,B}$ is not zero.

[diegoferigo]

After chatting with @traversaro, we realized that the previous logic was setting to zero the wrong $\dot{\boldsymbol{\nu}}$. The solution is to compute the bias link acceleration in the active representation (implemented in #127, that internally sets to zero the right $\dot{\boldsymbol{\nu}}$), and then convert back these accelerations to body-fixed representation to obtain the bias ${}_G \dot{\mathbf{h}}$.

I believe that we can unify somehow the complicated logic used in the functions to compute the bias accelerations of links and CoM, but this can be done in another PR. I want to move this forward since it is finally working fine.

[flferretti]

Thanks a lot Diego, LGTM!