resolve the issue of dependent contact constraints

[VenusPasandi]

We have two sets of contact constraints:

1. The holonomic constraints for contact vertices of robot feet,
2. The holonomic constraints added for modelling the closed loop chain kinematics.
   In some robot configurations, these constraints are dependent which results in some singularities in the contact and impact models. These singularities causes problems in QP solver used for computing the contact forces,

Possible solution https://git.robotology.eu/isaac/isaac-wp1/-/issues/646#note_25164.

[VenusPasandi]

According to https://arxiv.org/pdf/1711.08013.pdf, OSQP is robust for the positive definite objective function and dependent constraints. So, if we use OSQP as the QP solver, the only point that we should consider about dependent contact constraints is the mathematical computation of these constraints for example computing the inverse of a matrix that is rank deficient because of dependent contact constraints.