The (formerly) recommended number was "50 or less". My personal thinking is to just grab the half-dozen or so that seem most obvious -for-that-page-. In this way, not much time is lost, the keywords chosen are likely to actually be relevant and, when the search engine weighting pendulum swings back towards keywords again, you won't be scrambling to deal with a total lack of meta keywords across two hundred pages of content.
As "chief cook & bottle washer" for my blogs, I don't want to get caught with my pants down when that pendulum changes direction ... I already have my hands full. Better, I think, to take a couple minutes now each time I post an article than to have to read the articles themselves a year or two from now and conjure up the keywords then.
I do agree with OMZen, though, don't spend a lot of time on this. Just give it quick "lick and a promise" and move on to more important things (like the next page of great content or some wizard internal linking).
When adding meta keywords, I just scan my memory for words that actually appear in the article and that stand out as defining the topic (usually 4-6), type those and keep on a-moving. By doing it while the content is still fresh in my mind, it's a simple task.