Using the diode in the Casio projectors, the beam will appear close to a circle at the distance half way between where the fast and slow axis are in focus, but not along the whole length of the beam. So that is a white lie on their part. Anamorphic prisms won't do it, neglecting diffraction, because the magnification wouldn't be enough. Diffraction would pass through a small aperture near the diode and project the rectangle again at a distance. Cylindrical lenses probably won't because each one has to be at a different distance from the diode and at right angles to each other. A simple threaded tube won't allow that and the correct focus at the same time because it would need a half rotation to be in alignment again. The lenses would have different focal lengths, so focusing by turning that tube would adjust the focus of one lens much more than the other. The simplest way to do it would be to use a lower power diode, maybe 40-60 mW with a narrower stripe.
