That is nuts for light.
You'd need to physically switch light into different delay lines of optical fiber.
You could use a planar waveguide imaging splitter with an absorption modulator on each output. Hook up a variable length fiber to each output. Only allow light to pass through the channel with the appropriate delay. It would not be true phase modulation. It would have digitization error. You're not going to get that kind of variable delay electro-optically, or with any single channel, physical device.
Just looking at your reqs (30 microsecond delay at 10 megahertz) makes me think you don't really know what you need.