Instead of implicitly converting UINT64_MAX to operation destination
type, we explicitly set the correct max depending on type.

So instead of generating:

   uint8_t dst = (src0 + src1) < src0 ? (UINT64_MAX) : (src0 + src1);

we do:

   uint8_t dst = (src0 + src1) < src0 ? (UINT8_MAX) : (src0 + src1);

This clears following warnings during compilation:

  warning: implicit conversion from 'unsigned long' to 'uint1_t'
  warning: implicit conversion from 'unsigned long' to 'uint8_t'
  warning: implicit conversion from 'unsigned long' to 'uint16_t'
  warning: implicit conversion from 'unsigned long' to 'uint32_t'

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>