* rework application of deadzone
* Move application of deadzone outside octagon simulation.
* Scale result after applying deadzone back to the range ±[0, 1].
* Remove deadzone from octagon simulation, as a real N64 controller can
output output values in the full range.
* move application of sensitivity
The sensitivity is only for the real controller and doesn't really have
anything to do with simulating an octagon, so we move its application to
the processing of input from the real controller.
* use std::hypot instead of manual calculation
