The Openflexure microscope would be ideal for your graduate student workshop. I would try to build the bright field and fluorescent versions. It would be great if they were able to take one back to their lab afterwards, this is really what Open Science Hardware is about. The instructions for the fluorescence module are at present mainly in the Delta Stage, but the parts required for fluorescence are in the customisations page for the Microscope.
I don’t know Foldscope well, so I cannot really comment on your other suggestions for the school event. For option 3: if you are meaning to build an Openflexure device, a standard CMOS camera will not physically fit, and the sensor is not the optimal size. We strongly recommend the Raspberry Pi camera v2.1. The Openflexure optics are designed to make best use of the sensor size, and possibly more importantly the way in which the Pi camera is integrated into the Raspberry Pi hardware means that the Openflexure server has access to the video processing stream. We can implement live colour and flat-field correction and a fast auto-focus algorithm. For any other camera there is likely to be a strong colour cast. Compare the images before calibration https://openflexure.discourse.group/t/where-are-you-ofm-location-survey/771/44 and after calibration a couple of posts below that one https://openflexure.discourse.group/t/where-are-you-ofm-location-survey/771/47. Autofocus is also only available on the Openflexure Microscope with the Raspberry Pi camera v2.1. Without colour correction, flat field correction and autofocus it is not possible to use the motorisation to its fullest extent Automated Slide Scanning and Tiling.