also, mcp23018_init() needs fixing: `twi_stop()` needs to be at the end
of transmission blocks. i wouldn't think that would be necessary, but
it seems to be the only thing that'll make it work, and it also seems
consistent with the protocol diagram in the datasheet (lol, imagine
that), so i think that's what i'll have to do. not as though it matters
much i guess, since it's a single master / single slave system anyway, i
was just hoping not to release the bus till i was finished..