I went for an 8-bit mux/demux to multiplex the 8 data pins and lower 8 bits of the address because I thought it'd be faster, but I do have to delay quite a bit in the PIO program while I wait for the mux to switch. Given the very high speed that you can feed serial data into a PIO, a couple of high speed shift registers might have worked just as well and would have saved on GPIO pins which are a very precious resource.