Support irregular time-step in Runge-Kutta integration

[frothga]

A part which is queued to execute at a given interval may also receive (say) spike events which cause it to integrate the time between its last regular cycle and the time of the event. This means that when the next cycle time comes up, not all the parts necessarily have the same starting time for integration. When this occurs, then amount of integration time should be part-specific.

This is simple to implement with Euler integration, because it only considers last and current time. However, Runge-Kutta considers multiple time point between last and current. A possible solution is to pass current time and a value in [0,1] to indicate proportion of period.