D023 - Two pad models, on purpose

orbistoun and this project both hold a controller model, and the duplication is real. The part worth keeping separate is smaller than it looks, and it is the part that matters.

Its PadState is host-shaped floats, and its own documentation says why: "the guest-facing layout is unmeasured and this type deliberately does not guess at it." It describes what a person is doing with their hands, which is knowable, and leaves the conversion to whatever a title reads as a separate problem beginning with a measurement.

This project's Pad is unsigned bytes centred on 128, because that layout has since been measured against real hardware. Matching it means the payload does no arithmetic, and arithmetic in a payload is a thing that can be wrong somewhere nobody is looking.

Same question, different evidence, opposite answers - and both are right for their own project. Recorded so that neither gets "corrected" to match the other, which is what would otherwise happen the first time somebody noticed they disagree.

What genuinely is shared is the button positions and what is measured about them. A common home for those is worth having; the encoding is not part of it, and neither is the keyboard mapping - which is about a person and a keyboard rather than about a console.

Amended. This paragraph once listed tap-versus-hold among what is shared, which contradicts the last paragraph below, where it is refused for a reason that still holds. The last paragraph is the right one.

One premise has since expired, and the conclusion survives it. The quoted "unmeasured" was true when this was written; the button bits have since been measured and are credited in ACKNOWLEDGEMENTS.md. That does not merge the two models - it means there is a third thing, neither of them, which is the measured facts themselves. docs/PAD.md proposes where those should live.

Three of orbistoun's decisions were taken here and are credited in ACKNOWLEDGEMENTS.md. One was refused for a reason worth stating: its tap-versus-hold does not belong here. It exists because an emulator has to decide what a long press means. This project forwards absolute state sixty times a second and the target decides - so implementing it here would duplicate the target's own logic in the one place that cannot see the result.