padding: Clarity scope of nf_ito_*=0 disablement

For example, consider the (admittedly semantically nosensical)

   nf_ito_low                 0
   nf_ito_high                0
   nf_ito_low_reduced      9000
   nf_ito_high_reduced    14000

I don't think an implementation which knows that it is trying to do reduced padding ought to have to look at nf_ito_{low,high}. Clarify this by explicitly restating the 0-is-disabled in for the reduced values, in an effort to avoid the interpretation that nf_ito_[lh]=0 ought to mean to ignore nf_ito_*_reduced.