Fix duotone on slow IOPs

Also fix the DAQ Previously code assumed hkp cycle was always at the IOP_IO_RATE but this was a bad assumption and caused the dutotone data array to be filled in with bad data. Also, DACs were written only once per super-cycle, e.g. only once in a 16k cycle of a 16k IOP.

Note that a 16 k IOP has modle rate of 64k but with clock_div=4.