You can use package htmldoc, which is much easier to get working than pandoc.
It takes a list of html files and resolves the hyperlinks between them to
produce a single pdf with working links, and includes the images.
htmldoc $(find help/manual -name '*html' | sort) --outfile /tmp/out.pdf
Of course, it would help to know in what order to collate the html files. I
just sorted them on name.
Unfortunately, on my fedora 20 I got core dumps when I tried to put more
than half the audacity files through at the same time. I got to 440 pages of output though.
Perhaps just not enough memory. YMMV.
Remember for safety to prefer the real
audacity web site or your OS package manager
rather than sourceforge.