inverseDynamicsSolver should not realize stage Dynamics

[tanghekev]

The current inverseDynamicsSolver realizes stage Dynamics, which is not necessary. Stage Velocity should suffice, since this stage already contains all the information for the coriolis and gravitational forces. Therefore, the inverseDynamicsSolver can be made a lot faster.
See also my post on the OpenSim forum: https://simtk.org/forums/viewtopic.php?f=91&t=6292&view=unread&sid=c3815501f17b7ddc3653f1c146878bd7#unread

[sherm1]

Thanks, Kevin! We can continue the discussion here. From @aseth1:

I am already getting ~1000x real-time for an upper-extremity model - realizing to Dynamics is necessary if you want to account for other forces - say contact and ligaments (which I do want). The slowdown is the setup that the ID tool does and super slow reading of all the kinematics even if you want a small window.

Another issue we should consider is whether the Simbody 3.6 changes to speed up Dynamics stage (see PR simbody/simbody#435) make this less of a problem. (That change avoids calculating the Articulated Body Inertias until Acceleration stage.)

/cc @aseth1, @chrisdembia