This is mainly intended to make regression tests output filenames more
stable against changes of their contents and changes in lilypond-book,
and could also make generated snippets reuse a little easier (as it
makes output filenames controllable and predictable by the user).
This doesn't break the hashing system that saves lilypond processing
time.
See complete discussion at
http://lists.gnu.org/archive/html/lilypond-devel/2009-12/msg00636.html