Is it fair to my PhD student if I ask them to do "miscellaneous" work for a paper they're not going to be a coauthor of?

Preparing a graph from already-generated data, to me, doesn't merit co-authorship but would justify a mention in the Acknowledgements. However, something that takes two days out of a student's time (even if it would only take you 3-4 hours) seems like a lot to ask for just an acknowledgement.

How useful is learning the process? If it's legitimately training that's important to her future career, that might tip the balance toward asking her to do it. However, I think my default would be to do it myself.

One other part of the equation (probably and hopefully only a very small part) is that the student is female, based on the pronouns in the question, and female scientists are often expected (from unexamined assumptions) to disproportionately look after administrative work.

If you have to ask, then I suspect that at some level you know it's wrong.

I'm a PhD student, and I would find it extremely rude of my advisor if he behaved this way. It's not a student's job to perform menial tasks for his or her advisor.

I disagree with the "learning" benefit. She will have to learn it at some point, why not when writing her own paper. Learning it now won't take less time overall, in contrast, she will spend extra time drawing your figure.

I would not see a problem when you have a PhD student or postdoc who is good in preparing a special kind of figure, to a level way above yours. Then sure you may ask that student, and acknowledge the work (probably not by co-authorship, given there is no significant contribution - assuming that the paper could have been published with an ugly and harder-to-read figure).