[Fluxus] physics problems again

[gabor papp]

> I'm not so sure - I get "sanity check" errors sometimes that I think
> are due to normalization errors due to rounding issues.
we discussed this with dave and it seems that we would require the float
precision version of ode. double precision won't work unless we change
all fluxus matrices and vectors to double.

i changed the scons script so it autodetects the ode precision now. the
ODE_DOUBLE flag is no longer required.

what seems to be more worrisome that i tried fluxus with ode 0.11.1
single precision on osx and fedora and none of them worked. the osx
version crashes on the exploding sphere example, while the fedora
version kills the objects on collision. the transformation matrix 
elements of the objects become +nan.0. with ode 0.9 everything works fine.

best,
gabor