Am I breaking some kind of rule if I'm adding the arxiv link to the bibitem for an article that's published in a journal? I thought obviously not until I noticed that (link to arxiv) and (page number in journal) are quite mutually exclusive.
No. I think giving the arxiv link is fine. One alternative is to have two bibitem entries, one for the journal version and another for the arxiv version. I find it valuable to include the journal reference, because that shows that the article has been refereed (so its claims gain legitimacy). But I also see the value of including the arxiv version, particularly if the journal is hard to access.
Many copy editors will remove arXiv links to published papers, but usually they'll put them back in if you insist.
Technically, an arxiv version of a paper is different from a journal version. This can be a problem if you (for example) cite a particular theorem or lemma in a math paper that has a different number (or doesn't exist) in the arxiv version.
You didn't mention why you're adding the arxiv link. If you wish to add a link to a freely accessible document, then you could add a note in the text to that effect and cite the arxiv version in addition to the journal version. Otherwise, if you merely want a link to an online version, then you should be using the DOI link for the journal.