It's when an editor decides to run a manuscript past someone else to get their take on it. Obviously, it's a highly informal process, done (or not done) totally at the editor's discretion, unlike peer-reviewed journals. Likewise, there's no real reason to expect the reviewer to have any particular expertise in the subject material - they may be "peers" purely in the sense that they also have two arms and a head, like the author. In fact IIRC the "peers" that "reviewed" Design Inference were philosophers, not mathematicians, or biologists. Curious, since the book is all about math and biology...
a. The author is sometimes the reviewer of his own work.
b. Failing (a) if a reviewer reviews a portion of the book negatively, you just go and ask another reviewer.
c. The review process apparently consists not of the question, "Is this work valid science?", but "Will this book find a ready market?"