Generalized CMN notation model

[hanshoglund]

As part of the general clean-up needed for the backends. My reasoning:

• Currently we have two notation backends (Lilypond and MusicXML). Both contain a lot of similar code that has to do with translating the internal (nice, algebraic, transformable) model into something that more closely resembles a musical score. This involves dividing everything into bars, quantizing and rewriting articulation and dynamics into context-sensitive representations (i.e. beginning of ties, avoiding repetition of dynamic marks etc).
• This whole process would be easier if we had a separate representation that closely resembles a musical score. Sort of a generalization of MusicXML/Lilypond. The score models of Sibelius, Abjad, Lisp's CMN and PWGL are also similar (while Lilypond and music21 are closer to our model – being more logical than graphical).
• This would also make it much easier to add new notation backends (#25, #24, #171).

See also #278. Though my feeling is that we are approaching closure of that issue with the addition of aligned voices.

[hanshoglund]

Implemented in Music.Score.Export2* (to be moved). Closing.