refextract: improve author/affiliation extraction
- Added extra 'end of author section keyword'. Added a new 'bad unicode' character, and made sure to check for bad characters in both author lines and affiliation lines.
- Improve overall program flow of handling requests to extract authors or affiliations.
- Use both author matching and keyword matching when locating the top section of a document, in both author extraction and affiliation extraction modes. Plus other improvements.
- Enrich keyword kb.
- Extend author name numeration to include brackets.
- Update usage() to include '--affiliations' option.
- Add a comment explaining why a top section would not be found, and remove some redundant spaces between statements.