To be properly balanced (and not to fall backwards or forwards),
shouldn't the two torques (referred to the midfoot) be equal and
No. In order for a body to be balanced, the sum of all external torques, including both pure torques and torques resulting from coupled forces, must be zero.
In the case of a human squatting, there are no external pure torques applied to the body, and there any only three forces acting, those being gravity acting on the lifter's body, gravity acting on the bar, and the ground reaction force of the floor pushing up on the lifter's feet. These three forces can be couples into two moments - the moment between the lifter's weight force and a part of the ground reaction force, with a moment arm equal to the horizontal distance from the lifter's centre of mass to their midfoot, and a moment between the barbell's weight force and the other part of the ground reaction force, with a moment arm equal to the horizontal distance from the bar to the lifter's midfoot. These two must cancel each other out in order for the lifter to remain in balance, which is normally achieved by the lifter's centre of mass being behind the midfoot, and the barbell being slightly in front of the midfoot.
Both the hip and knee torques are internal torques, and so they already balance themselves out when considering the body as a whole, and do not affect whether the whole body will rotate. The reason that they balance themselves is that the hip torque actually consists of (when viewed as in the above diagram) an anticlockwise torque applied to the torso by the thighs, and an equal clockwise torque applied to the thighs by the torso, while the knee torque consists of a clockwise torque applies to the thighs by the shins, and an anticlockwise torque applied to the shins by the thighs.
Is it important [...] that the two torques [...] are not too high? These analyzes stress the fact that the knee torque may be greater (bad) or lower (good) accordingly to the squat variation we choose (front, low bar, high bar). But it seems a non-sense to me, as we want to gradually increase the load as much as possible whilst mantaining proper form.
Yes, your intuition is correct here. It is a common mistake in biomechanical analysis of lifts to try to minimise knee torque, for fear of causing knee injury. This is where the myth of squatting with your knees moving beyond your toes being harmful came from. But creating stress on the knee is actually the entire purpose of leg strengthening exercises, and is the mechanism through which growth of the quadriceps is stimulated.
The consequence of attempting a lift where knee torque demands are greater than what your body can provide is not injury, it is merely that the lift will be failed.
Which are the "Knees stabilizers" the website refers to and which is their role?
As a rule of thumb, I would suggest that any time anyone talks about "stabiliser muscles", or worse still, the ubiquitously used "small stabilising muscles", without being able to actually name those muscles, that person probably doesn't know what they're talking about, and should be ignored.
There are two definitions of "stabiliser muscles": They are either muscles which contract isometrically to maintain posture or fixate a joint (such as the front deltoid during a bicep curl, or the trunk musculature during a squat), or the muscles responsible for keeping the load in balance in a free-weight lift.
I'll use the latter definition, because including any isometrically active muscle as a stabiliser means that the back is a stabiliser for the deadlift, or that the rectus abdominis is a stabiliser for the plank, rather than them being the primary targets of those exercises.
In the squat, the only challenge to balance is the weight tipping too far forward or back, and it is actually the prime movers of the exercise (the calf, quadricep, and gluteals) that are responsible for stability.
Some exercises do have distinct stabiliser muscles though, for example the bench press primarily works the pectoralis and triceps muscles, but uses the latissimus dorsi and deltoid muscles for stability, where the lats activate to bring the bar back if it shifts up over the face, or the deltoids will activate if the bar shifts down too low on the chest.