A poster on StackOverflow suggests that the (huge, 22GB) RDFs can be zgrepped. I haven’t gotten that far myself.
I snagged an old dump from October, 2012, weighing in just short of 54 MB. I suppose it will be adequate to replicate the book’s exercises.
That file doesn’t include performance.tsv, needed for the polyglot chapter. But I found two relevant GitHub repos. They’re designed for use with Vagrant and VirtualBox. But they can be downloaded as Zip files. The .tsv files can then be located and stripped out. (Or, Vagrant can be installed and the chapter databases can be set up automagically.) See codingbynumbers, the more complete of the two, and paulspencerwillams.