Instead of using xslt, we now process the documentation in code. This
allows us to fix mismatches from e.g. invalid parameter names that
sometimes creep in the specs.
Improved the output for the doc processor: MathML equations are replaced by their textual representation and whitespace better matches function declarations.
Added new configuration settings for the doc processor and the license.
Removed stale debugging code.
License.txt is now a file, not a resource.