It's not just a matter of one being primarily for creating vector based graphics and the other being primarily for creating raster images (although that certainly matters). It's also a matter of one being an "object based" program and the other not.
A whole-document project file is not just one image or one vector path. It's a stacked and arranged collection of raster objects, vector objects, and text objects. Some of those objects may be printers marks, die cut paths, spot color separations for varnishes, and on and on. Object-based replication just makes sense. In a typical business card press-sheet, for example, you may have a gang-up of 10 instances of the same card, with trim marks for all. Programmatically replicating one instance in a variable way makes more sense than actually stepping/repeating an array of pixels.
You should also bear in mind that much of Photoshop's "vector" capability is smoke-and-mirrors. Shape Layers ar actually selections of pixels for painting the pixels they enclose. The result is just colored pixels, not printer commands for grads, etc. A Photoshop Layer is a very different construct from an Illustrator Layer. Consider, for example, creating in Photoshop a stack of individual vector objects which have multiple stroke and fill attributes and which look like a single illustration when viewed together, but which are still independently selectable and draggable as objects, not as layers. Select two objects from Layer 1 and three objects from Layer 3 and transform them at the same time. A Layer in Photoshop is really another whole raster image, just as large as the whole Photoshop document.
Still don't believe it? Create a "vector" Shape Layer in Photoshop. Position that "object" in the upper left corner of your layout. Create another one and position it in the lower right corner of your layout. What's the PPI resolution of your document? I'll bet it's the conventional-wisdom 300, right? Okay. Give the "object" in the lower right a drop shadow. Now, bear in mind, there's absolutely no reason that drop shadow needs to be sampled at 300 PPI. 150 PPI is a gracious plenty. So give just that drop shadow a PPI of 150. Can you do that?
And what of all that "empty" space between the two "objects"? Is it really empty? Really? No, it's filled with thousands of absolutely useless white pixels (or grayscale pixels in an alpha channel).
Generally speaking, if you are assembling a whole document, you are working in an object-oriented "assembly program" in which you combine the assets from vector programs, the assets from raster programs, and native text assets. Conventional wisdom calls that "assembly program" a "page-layout" program. So conventional wisdom supposes the need for a raster (painting) program, a vector (drawing) program, and an assembly program. The fact is, a drawing program comes much closer to being suitable as a "page-layout" program than does a raster imaging program. Vector drawing programs and page-layout programs share the same kind of object-based nature. The significant differences between them is that the features of a page-layout program are more geared for semi-automation of repetitive layouts (master pages, etc.), high page counts, and long threaded text handling. As long as you are not talking about long "bookish" documents, a drawing program works perfectly fine as an assembly program for short-page-count whole-document projects (usually even better, truth be told). The same is just not true for a raster imaging program, because when it comes down to it, a raster imaging program is still all about creating one raster image.
JET