There is a huge gulf between what would happen in an ideal world and what the norms are in this case.
"Ideally", you'd get in touch with the author(s), explain that you don't see what their paper has to do with yours; they'd explain why they think it is relevant or agree that it's not, and modify the paper accordingly. (Almost everything is online, so modification after publication is a possibility.)
In the actual world, citations are of benefit to you, even if they are stupid. Journals are mostly not set up to remove citations easily. No one will check, and if they do check, the detriment will be to the citer, not the cited. So you "shouldn't" do anything about it, and the author would probably be quite surprised if you did (especially if you weren't discreet about it). If you really feel like re-calculating your h-index with that paper removed, go for it. But this sort of thing happens all the time (I think all of my papers with over about 50 citations have been cited stupidly at least once), so you're free to just consider it part of the measurement error inherent in looking at citations.
Incidentally, the ideal isn't necessarily the pragmatic ideal. Doing anything important on the basis of small differences in numbers of citations is fraught with error even if all citations are sensible ones. There's a reasonable argument to be made that you shouldn't bother unless the paper is in your field and is citing you in support of something that your paper showed the converse of. Getting your work exactly backwards to advance their own idea isn't doing you or them any favors, so you should try to work that out.