My inclination, if you weren't worried about harnessing energy, would be to use some big MOSFETs to switch current to some resistors or light bulbs and then use a microcontroller to control the MOSFETs. Light bulbs are cheaper than power resistors, and they're designed to dissipate power (given a suitable enclosure), but their resistance varies substantially with temperature, which could complicate your logic.
The basic idea would be to pulse-width-modulate the MOSFETs to vary the load based upon what the processor determines is appropriate. To avoid having massive flyback pulses kill the MOSFETs, I would suggest having a number of parallel resistor-plus-MOSFET combinations, plus a resistor which is "always in", and a circuit to turn a MOSFET full on with a moderate-value resistor if the voltage gets too high. If you switch from having 10 ohms of effective resistance to having 25, that will cause an instant momentary 2.5x increase in voltage; the MOSFETs should be chosen to allow for that.
I would expect that monitoring and controlling the effective mechanical resistance is going to be easier if one can switch the MOSFETs and resistors in such a way that you never hit a flyback clamping point. If your allowable flyback voltage is 5x the voltage produced by cranking, it may be good to have two MOSFETs and resistors per decade (e.g. values of 1, 3.3, 10, 33, and 100 ohms). To produce effective resistances between 33 and 100 ohms, turn the 100 ohm resistor on and modulate the 33-ohm resistor. This should limit the flyback voltages to 3x the voltage produced by cranking.