There are two ways to look at it, a mathematical way or an alternative, intuitive way.
The alternative way can be to look at F as an alternative S with 0 interest rate discounting because we still have the cash (minus a small posted margin, and ignoring this) which earns the interest rate. So for the F’s value itself every day’s time value of money effect is zero and the daily mark-to-market makes a PNL transfer from the cash posted. More specifically , for S we need to use discounting to arrive at F price, but when F itself is the new spot, we don’t need further discount, as we don’t pay the value of F.
In the traditional way, $(C_s - P_s) = D (F-K)$ is correct when both $C_s$ and $P_s$ are options on Spot. But in the case of CME options, the options are all options on futures.
Let $C_f, P_f$ be options on futures. At option expiry, the option gets converted to a future not a spot , which has a discounting factor vs spot.
Making some assumptions on the options expiry date (which in practice is on or before futures expiry date, and also ignoring a delivery period which causes a further mismatch between futures expiry and spot conversion):
Similar to $S = DF$, one can write $C_s - P_s = D(C_f-P_f)$.
So it becomes:
$D(C_f-P_f) = D(F-K)$, or $C_f -P_f = F-K$.
However, in the case Equity indices , usually options on the Index (e.g. on CBOE, or on Eurex) are more popular than options on futures. For these options the original wiki formula would apply .