We use the new merge feature of the link rewriter to suffix each xml:id
in the corpus with a unique string generated from the respective
object's URI. This not only makes the corpus export generate well-formed
documents but it also fixes EPUB generation for cases with duplicate IDs
from different objects.