[core] Wrong computation of data.f in computeExtraTerms

[duburcqa]

In Jiminy, data.f is expected to store the internal forces applied on each joint under all external forces on the whole system including the gravity field. This quantity is actually a by-product of RNEA when considering all external forces including contact forces.

For now, data.f store the total external forces applied on each body including joint effort and reciprocal forces (namely the forces that would be involved in Newtonian formulation of dynamics), eg 0.0 if in static posture.