Sum forces for each clump (and torques for "spherical" integration)

Following r2637.

Clumps currently get accelerations from clump members in Newton.
It would be better to define total force and total torque for each clump with addForce/addTorque. Two reasons are :
- It would not require divisions as f/mass and t/inertia for each other clump member, instead there would be only one division of the total force/torque.
- It would give a correct meaning to getForce(clumpId), so that scripts can get actual forces without the need to iterate on members, unbalancedForce would apply for clumps, etc.

Aspherical rotations are already almost on this line, computing the total torque "M" first, then acceleration.
M is only a temporary variable however, we could assign it to clump force before forgetting it.