This seems like a legitimate use case for an FPGA in every PC. Implement a set of expensive algorithms in Verilog, put it up on Github and set up some slick distribution mechanism that updates the logic when needed without bothering non-technical users with it. This would prevent the crypto from using up too much general purpose processing resources, cause it to be acceptably fast and provide a mechanism for updating the algorithms when flaws are found or new algorithms are introduced. Downside is that it would probably take up to a decade to actually be adopted by the majority of PC users, but then again, the majority of PC users will probably not care too much.