Having thought about this a bit more, and reflecting on comments, I thought I'd add a second answer rather than heavy editing on my old one.
I think there are three slightly different matters that are being talked about here.
1. Referring to a user whose name or icon suggests a gender
In an ideal world we would not make assumptions based on these things, but in reality, at least for now, this is going to happen, and the number of people for whom it will be wrong, as a proportion of the population, is very small. Except in the case where somebody is being deliberately misgendered (which would come under anti-harassment rules), I think this is something to leave alone.
2. Referring to a user whose gender is not suggested by name or icon
Here, users should make every effort to refer to that user by gender-neutral pronouns. It must be pretty offputting for a woman to post here under a neutral name (for whatever reason) and be assumed to be a man - or, probably less common, vice versa. It is not appropriate to make assumptions, which may tend to reinforce stereotypes.
I believe that this should be noted in the help, and people should make an effort to remind askers and answers of this policy in comments.
3. Referring to a generic person
I mean here when somebody is not talking about a specific SE user, but is referring to, for example, "a student" or "a postdoc". Here I think that it is really important to remain gender-neutral, lest we reinforce the idea that all researchers in $field are $gender.
I think this is by far the most important of the three items here. I think that correcting gender-specific terms used in this sense should, in itself, be grounds for editing a post. If there is argument over this on a specific post, then preferring gender-neutrality in generic people should be the guidance for moderators.