I'm pretty sure the ATMega8 has enough I/O pins to make that work -- my understanding is that the USB communication is just done by bit-banging, so the code is pretty simple. I can check that for you, and it might be the case that a surface-mount version, e.g. QFP, might have more I/O lines than the DIP version. Also, if the DIP version doesn't have enough I/O lines by itself, it'd probably be possible to multiplex it.

Let me look into it and get back to you; if I haven't done so in about a week, kick me about it.