Split scriptconv stage from parseconv.
Part 2 of redesign for parseconv & generator (planned in 2013). Not exactly as originally thought of, but as a basic preparation, split parser output representation from script data representation which is later converted to the internal audio generation "program". The parser is still mostly the same, with a very small simplification. The old parseconv has become scriptconv, with minor adjustments for input data changes. New parseconv mostly copies parser output, but leaves flat per-event operator list. Todo: More refactoring, further redesign (for which the door is now open). No functional change. Increase version to v0.3.3. Change log: * 2019-10-16: Placeholder for separating stages. * 2019-10-23: Split parseconv module into parseconv and scriptconv. * 2019-10-25: Minimal cleanup. * 2019-10-26: Fill in more, still not functional. * 2019-10-27: Make functional. Set version to v0.3.3. * 2019-10-28: Fix 'test-builder' build.