If you compile to pdf, there are a couple of commercial programs (Pdf Index Generator is one) which let you create an index, but it’s really more of a concordance. You can tweak the automatically generated output, but that might take a while, especially if you have a long work. The problem is removing common words and your work will have its own common words. In a bio of Napoleon, listing all the pages where the word “Napoleon” appears won’t be very helpful. In my experience, such a program only gets you halfway there. But in many cases, a concordance is what is called for.
You can Compile to LaTeX (or text for use in LaTeX) from Scrivener. That way, creating an index for LaTeX from inside Scrivener is technically easy. As you are writing your content, wherever you feel a reference should be included, you add “\index{my reference}” and when you generate a pdf in LaTeX (or a text file for use in LaTeX) the references and page numbers will all match. Unlike a concordance, the advantage is that the index is not limited to words that are in the text itself, you can use whatever language you like in its creation. There are different ways to style an index, but that is outside the scope of your question.
If your budget allows, you can use a professional indexer but you will still have to append their output to your Scrivener-compiled pdf.
E-pub indices are another matter. Some publishers just include the plain text of the index with a disclaimer saying that the page numbers no longer correspond to the text (ePub 2) though they should in ePub 3. Others convert the index page numbers to URL’s (after all, an ePub is just a web page) which is more useful.
Your note about nested keywords is interesting, what would this look like?
Too bad that “pure” Markdown doesn’t have index functionality (it has footnotes, though). I don’t know whether Multimarkdown has this feature built-in or not.