The twist is distributed linearly from the root to the tip.  I'm not sure if that is ideal, but it is probably good enough, and it was the most simple way I could think of to do it.

The tangent of the angle of attack (slope, i.e. opposite-over-adjacent lengths of the right triangle with the chord as the hypotenuse) is inversely proportional to the radius (= constant * 1/r).  Not to hard to compute.

A linear twist is far from ideal - especially near the base - but is still a lot better than no twist at all.  Much less of the blade is stealing power from other parts and using it to pump air instead of pulling its own power from the air.

If I were doing one I'd let the trailing edge go "off the back" of the workpiece near the hub.  Then I'd cut one or more shorter pieces as necessary to fill in the missing part and laminate them.

With a CNC machine there should be no problem getting the layers to line up.  Also:  You can cut alignment groves and ridges in the mating surfaces to make them self-aligning when you stack them - or cut slots in both for an strip of wood to make an alignment key to do the same thing.  Then just spread glue, stack them up, and clamp them (or put weights on them) until they dry.  (You can also cut matching bed pieces to stack but NOT glue on, to give you flat surfaces for your clamps or weights.)

An earlier article discussed good choices of glue for laminating props.