You could do it optically.
If you have a series of plastic rods of different lengths, and a small light source, you may find that a sensor (phototransistor) bonded to the top of the rod will detect a difference when the end of the rod is immersed in the fuel (you would have to do some testing).
This method is used in a device I have to detect the levels of solution in a darkened tank (in this case ambient light is the light source) I think it relies on suppressing the reflection from the polished end of the rod and would also depend on the indexes of refraction of the plastic tube and the liquid.
Another possibility (and a pretty clever one I think) is to use the volume of air above the tank to control the frequency of an oscillator. I'll have to find a reference for this, but I saw some circuits which placed a speaker in a variable size enclosure (think accordion) and used feedback to get an oscillator to oscillate at the resonant frequency. The beauty of this is that it relies more on the volume of air in the tank than where it is located. The problem with this is that the sound of the engine may make it impractical.
Hmmm, I can't quickly google something useful...