#5 Bug using noise multiplied by a parameter

[Romain Brette]

When using "sigma*xi" in an equation where sigma is a parameter (or a variable), it will not work (sigma will be replaced by its unit value). This is because only additive noise is currently implemented, as a separate mechanism on top of the state updater.

One way to fix it:
For nonlinear updates, merge the two updaters (deterministic + noise) into one stochastic updater, e.g. by replacing xi by randn (for Euler). This is not completly trivial though, for second order ones the mechanism is more complicated.

Another problem is that when noise is not additive, one has to choose between Ito and Stratonovich interpretation (I would go for the latter which is more physical).