[ODE] Improved QuickStep stability...?

[Adam D. Moss]

I've just checked in another minor tweak I was testing earlier;
it re-random-permutes every 8 iterations instead of every iteration.
It doesn't much negatively impact stability (most of the random-
reordering stability gain comes from doing it at all; doing
it multiple times per step is a relatively minor improvement) but it
does hide most of the slowdown seen with large numbers of iterations
on large (I guess cache-busting) simulations with
RANDOMLY_REORDER_CONSTRAINTS enabled.

--Adam
-- 
Adam D. Moss   . ,,^^   adam at gimp.org   http://www.foxbox.org/   co:3