The reason is the definition of $\hat{r}$:
$\hat r$ is a unit vector which is pointing from the other charge to the charge itself.
The vector version of Coulomb's law is:
$$\vec F_1=k\frac{q_1q_2}{r^2}\hat r_{21}$$
Note the difference in notation from your expression: $\vec F_1$ is the force felt by charge 1. $\hat r_{21}$ is the unit vector from charge 2 towards 1.
Now, like charges (same sign) repel, so the force will point in the same direction as $\hat r_{21}$ - in other words, away from the other charge.
For opposite signs, the $q_1q_2$ term will be negative, so the force will turn around. The force will point in the opposite direction as $\hat r_{21}$, which is towards the other charge. Which it also should, since they attract.
So, you can't add absolute values here. Then the formula would only be correct for like charges. Their signs take care of the correct direction of the force in this formula.
The key point is that the unit vector $\hat r$ doesn't point from the charge itself to the other, but rather from the other charge to the charge itself. Be always very clear about such definitions of each parameter.
When only the magnitude is needed, we do not care about direction. Which means, we do not care about any signs, since the signs only take care of direction. Therefor your expression no. 1) above has absolute values, so that any possible signs are removed. The magnitude is then always positive, and we must just remember that directions cannot be seen here.