The move to Nehalem will be in many ways as big as the original move to the Intel platform. The new multi-CPU interconnects and integrated memory controllers will have a large impact on performance, just as they did for AMD when they incorporated them with the Opteron platform.
So that implies Apple will want to shout about it, which implies they'll use either MacWorld or the WWDC. Since programming for Nehalems will be essentially just like programming for Penryns, launching the product at WWDC is unlikely. So that leaves MacWorld.
Another reason why launching at MacWorld makes sense is that while the Mac Pro will likely be the first Macintosh to use it (unless Apple decides to move to desktop chips for the iMac), it will show up down the road on the iMac, MacBook and MacBook Pro (and perhaps the Mini). So Steve can announce it in the Mac Pro, show how amazing it is, and then note that within a year you'll see it on everything else in the Mac line-up.
The alternative is that Apple decides to stick with FB-DIMMs and perhaps move from two CPUs to four (to maximize Grand Central) for the Mac Pro. At that point, then MacWorld 2010 makes sense as the launch for the entire Macintosh line - tower, all-in-one and mobile - with Nehalem, since Intel will have all the CPUs and chipsets out by then.
A possible fly in the ointment is how fast OS X 10.6 arrives, especially Grand Central. If they can get it out by MacWorld 2009, then that supports the Mac Pro release then. If they cannot, Apple might wait until WWDC 2009 and launch the Mac Pro then with 10.6. WWDC will likely be too early for Clarksfield and Auburndale - the mobile Nehalem platforms - but close enough that Apple could show some iMacs and MacBook(Pros) with engineering sample hardware. However, this will likely kill existing sales...