[ODE] More save/restore issues

[Eric Froemling]

In my never-ending quest to get numerically-perfect save/restore 
working with ODE, i've come across a few more issues - I thought I'd 
throw them out and see if anyone had any suggestions or comments in 
case im missing something or misintepreting something...

The first problem applies to simple or hash spaces - i havnt tested 
with quadtree but at a glance it looks to be ok.
When a dxGeom is added to a space, it is placed on a list in the space. 
  The space's list is sorted such that "dirty" geoms (ones needing 
bounding boxes updated or whatnot) are first in the list.  Thus when a 
geom is dirtied, it is re-added to the list at the beginning.  Thus 
when i run my simulation from frame 0 up to frame x, at frame x the 
geoms are all shuffled around from being dirtied many times up to that 
point, but if i recreate the simulation to start at frame x, the geoms 
appear on the list in the order they were just added.  This means the 
order they are returned in for collisions and other such things will be 
totally different and in the end the sim doesn't come out the same.

As a quick workaround for myself, I removed the reordering code in 
dxSpace::dirty() and modified dxSimpleSpace::cleanGeoms() and 
dxHashSpace::cleanGeoms() to scan through all geoms instead of
stopping when it finds a non-dirty one. (since they are no longer 
grouped at the beginning of the list)  I realize this reduces 
efficiency a bit, but it fixes the problem for me.  It looks like the 
quadtree space keeps a seperate list for dirty geoms - maybe that would 
be a good way to go for hash and simple?

The second bug is a bit simpler - I found that retrieving aMotor angles 
for an aMotor constraint i've just created gave incorrect values - Upon 
examination, I noticed that dJointGetAMotorAngle simply returns 
joint->angle[anum], which is computed during amotorGetInfo1, so the 
returned values are accurate only after a simulation step has been run. 
  To fix this I just added the 
amotorComputeGlobalAxes()/amotorComputeEulerAngles() call into 
dJointGetAMotorAngle().

-Eric