That's the Cypress FX2LP USB microcontroller (I recognize it because I use it myself). If you're using the Hi-Speed USB transceiver, then you should really go with a 4-layer board. Without that ground plane right below the top layer, it will be near impossible to get the 90 ohm differential impedance that you want for the USB D+/D- lines.
http://www.cypress.com/?id=4&rID=34128 flat out states that 4 layers is required. It also states that controlled impedance is required, but in my experience you can usually get away without it, so long as you carefully research your fab's typical stack-up and work out the right width, space, and height.
http://www.cypress.com/?docID=25406 also provides more info on calculating the width, space, and height for the D+ and D- lines.
4-layers isn't that much more expensive; Advanced Circuits has a 66 each deal for 4 layer boards that I use quite often for projects that use that very chip, as opposed to the 33 each deal for 2 layers.
In regards to your actual question...use plenty of bypass caps, as close to the pins as possible. If you split the bottom layer to have VCC and GND, don't have a trace cross the split on the top layer. Keep all high-speed signals on the top layer because the via inductance can kill what fragile signal integrity a 2-layer board has.
I think you're on the right path, a couple of notes,
1) With a signal trace between two planes, the return current will split between the two planes, even if one of the planes is split. The return current cannot "see the future" and decide ahead of time which plane to return on. It will return above and below the trace until it sees the split at which point is says "oh crap!" and pays you back by possibly causing you to fail FCC testing. So you want to avoid running traces over plane splits even if another adjacent plane is not split. You can deal with splits with capacitors and such but this type of solution is less than ideal. I'd focus on always avoiding running a trace over a plane split on an adjacent plane.
2) Wide return paths on DC signals don't really matter.
3) You asked about two signal layers sharing the same plane. Usually, this is not a big deal if done properly. What many people do is use one of the layers as a "horizontal" signal layer and the other as a "vertical" signal layer so the return currents are orthogonal to each other. It is very common to route two signal layers for each plane, and use this horizontal/vertical technique. The most important thing to remember is to not change reference planes. Your setup could be a little tricky because going from the bottom layer to the 4th layer adds another return plane. More typical 6 layer boards are
1)ASignalHor 2)GND 3)ASignalVer 4)BSignalHor 5)POWER 6)BSignalVer
If you need smaller additional planes, like under the micro, these would usually be placed as an island on one of the signal layers. If you need to use more power planes, you might want to think about going to 10+ layers.
4) Plane spacing is important, and can have huge impact on performance, so you should specify this to the board house. If you take the example 6 layer stackup I mentioned above, spacing of .005 .005 .040 .005 .005 (instead of standard stackup with equal distance between layers) can make an order of magnitude improvement. It keeps the signal layers close to their reference plane (smaller loops).
Best Answer
I've done this before, and it does require some precautions that you set those pins to inputs on the MCU. Other than that, most MCU pins can withstand short circuit for a short while.. so I think it is survivable if you set them up incorrectly during development.
Alternatively what you could do, is add soldermask to the pads that you use for snaking traces like this. That way the QFP pin doesn't get soldered to the board.
However, not sure how well this trick will hold up if you plan on mass producing the board.