Stochastic equations can be falsely identified as multiplicative

[mstimberg]

The following code will not run, because Brian considers the noise to be multiplicative:

ta = TimedArray([0, 1], dt=defaultclock.dt*10)
group = NeuronGroup(1, 'dv/dt = ta(t)*xi*(5*ms)**-0.5 :1', method='euler')

For the exact state updater, we already deal with this situation by checking whether ta is constant over a time step, we should do the same for stochastic equations.

The workaround is straight-forward, you can select a state updater that supports multiplicative noise (e.g. heun)

First reported by Martin Bertele on the mailing list.