Consolidating multiple projects for mastering - what's 'normal'?

Stems are used to give the mastering engineer more flexibility, essentially providing him with a way to modify your mix on a macro level. MEs love this, because they are so often expected to fix things that should have been fixed in the mix.
 
But since you are both the mixing and mastering engineer you don't need that capability. Go with your original plan, keep it simple and export 32-bit stereo wave files for the mastering project. You can still apply separate plugins to each song if necessary - just load the imported pre-masters into separate tracks in the mastering project.
 
I like to use a separate audio track for each song, which lets my apply EQ and compression as needed. First step is to identify the hottest track and use it as the reference for all the others, turning them up to approximately match its volume. Do this by ear rather than meters and don't worry if turning them up results in scary peak values. If one track is too peaky to turn up to match the reference, apply some compression or (gentle) limiting.
 
Don't overlook the biggest advantage to using a separate mastering project, which is adjusting the gaps between songs. Listen to how each flows into the next - you might even alter the song sequence if a transition is too jarring. At the end of one song, keep counting the beat in your head while waiting for the subsequent song to begin; if it's a similar tempo then you'll want it to start on that imaginary downbeat. Also consider crossfades instead of silence, which can be very effective when the songs are in the same key, or when the first song ends with a fadeout.