I don't think someone is going to implement a conversor from SynFacilsyn XML files to Pascal units, because SynFacilSyn can work as faster as a hardcoded highlighter. Not worth it.
And is someone is animated to do something like this, probably will choose a different syntax, considering the SynFacilSyn XML syntax file is optimized to define a fast lexer, more than be flexible.
IMHO SynEdit architecture is some limited and has to be redesigned or replaced. Modern IDE requires powerful editor machines, with features like plugin support, graphic interaction, multi-languages support, solid API for completion, highlighters and codetools.