Natural selection cares only about if the species is able to reproduce successfully.
Inhibitory mechanisms to aggression
Natural selection promotes genes associated with violence and aggression
Aggression can increase evolutionary fitness
There exists specific, recurring situations where aggression increases fitness
Buss, David and Todd Shackelford. Human aggression in evolutionary psychological perspective
In humans, as most people who are normal are able to reproduce, its only a matter of finding a mate for them. Some people might prefer that their partner is aggressive while some may want a calm partner. As of this trait passing down to offspring, I would think that these traits are learned in life and genes play a minor role in whether the child is aggressive or not.
In-group aggression differs from out-group aggression. It is characteristically less lethal than out-group aggression and can act to jeopardise the survival of the group.
Further, aggression is a high energy, high risk trait and following N2 need only be promoted by the particular circumstances in which it does enhance reproductive success i.e. resource scarcity. Otherwise may be negatively correlated to fitness.
Multiple species including ours have evolved inhibitory (seretogenic) mechanisms to aggression.
Firstly this suggests that aggression is not unreservedly correlated to fitness and secondly, demonstrates that genes associated with a lack of aggression are (also) promoted by natural selection.
There are multiple definitions that exist of 'fitness' in the evolutionary context, as explored in https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CCQQFjABahUKEwiFsd-AjPDHAhUHlogKHTdmDfc&url=https%3A%2F%2Foid.wharton.upenn.edu%2Ffiles%2F%3Fwhdmsaction%3Dpublic%3Amain.file%26fileID%3D5444&usg=AFQjCNGp1KQpRoCbkAehYLZuJtQvyMSCiA&bvm=bv.102537793,d.cGU&cad=rja.
In this graph, a gene associated with fitness is one that leads to a greater number of successful offspring. From the above link, a more explicit measure of fitness: "Waddington's characterization of fitness as 'the capacity to contribute offspring to the next generation' (1957, p. 109) is a standard conception of fitness which is explicated by defining fitness, in the absolute sense, as an expectation. Crow & Kimura define it as follows: We define fitness, or selective value, as the expected number of progeny per parent. [...] (Crow & Kimura, 1970, p. 178)."
This article puts forward seven different "leading candidates for adaptive problems to which aggression might be an evolved solution." Multiple nodes into n2 were adapted from this article, with some additions.