Well, as distributed you only need to run two script files, and these could be merged into one. I'm not sure why Ars Digita hasn't done so, probably is due to the fact that the geospatial stuff was added for a single client and as of 3.1 at least still wasn't cleanly integrated into the toolkit.
Now that we have installation instructions, this question's obsolete as the instructions tell you how to run the two install scripts.
The problem with pg_dump output is that it is kinda unreadable, spewed forth in a legal but obtuse format that includes a lot of redundant information (emitted to minimize the amount of bookeeping work done by pg_dump, it is easier to dump everything that other tables might need to reference rather than just the stuff actually used).
So we'd need to distribute the "clean", human-formatted and commented datamodel anyway.
Because of these reasons, I don't see any benefit to distributing a dump rather than the two initialization scripts - now that we have documentation!
