Split kinematics functions (forward, backward, reconstrution) into their actual input/output relation

Right now, the different kinematics ODEs are set up in a "nested" way i.e., reconstruction is Q, r, \eta, forward is Q, r, \eta, \dot{\eta}, and backward is Q, r, \eta, , \dot{\eta}, \Lambda, Q_{a}.

Since the backward kinematics are actually only ever needed to determine the F_{0} = \Lambda (0) and Q_{0} = Q_{a} (0), it would make more sense to also only return these respective components from these functions. As such, these functions are 1:1 implements of their rigid joint counterparts and require no special treatment (even though one could argue that we could further split the result of backward into backward_force and backward_actuation to be fully compatible with rigid joints, however, I believe that a function like [F0, Qa] = backward(joint, Q1, r1, eta1, eta1dot, q, qdot, qddot, F1) is likely as good as two functions F0 = backward_force(joint, Q1, r1, eta1, eta1dot, q, qdot, qddot, F1) and Qa = backward_actuation(joint, Q1, r1, eta1, eta1dot, F1, q, qdot, qddot).

In the end, we would have

[Q, r, eta]         = reconstruction(joint, Q0, r0, eta0, q, qdot, qddot);
[Q, r, eta, etadot] =        forward(joint, Q0, r0, eta0, q, qdot, qddot);

[F0, Qa] = backward(joint, Q1, r1, eta1,          q, qdot, qddot);
[F0, Qa] =      idm(joint, Q0, r0, eta0, eta0dot, q, qdot, qddot, F1);

which, I believe, can provide the same syntax for rigid joints and flexible/soft joints that may be used directly inside the RIDM/RNEA algorithm.