The Internet search-engine giant wants to scan all 7 million volumes in the U-M collection, a project estimated to take at least six years.