Great idea, but I think Google already uses this data.
Google Toolbar and now Chrome report this data back to Google, and most search pros believe "serp bounce back" and "time on site" are key signals Google uses.
PageRank and DwellRank are not either-or choices.
Here's my theory: Google uses PageRank to decide what pages to "try out" for a query (i.e. display a page in the SERP for a sampling of queries). If the page gets clicks AND has good "DwellRank" then it gets progressively better and better rankings. If a new page enters that beats it, it falls.
This approach is very Googly -- they love to test. They love to decide if product features are good or not by giving them a sampling of traffic. It would be insane of them not to extend this approach to search.
So the upshot is, use "PageRank" to decide which pages deserve an audition, and use "DwellRank" to decide the winners.
Since 40% of the clicks go to #1, 10% to #2, 8% to #3 etc,Google can audition pages using DwellRank without affecting the experience of the majority of their users.