refextract: improvements to author recognition

• support for the unicode character 's' with a tilde
• support for volume numbers separated with a hyphen
• making sure that the first author in an author group, if starting with an 'A' initial, must have a full-stop after that first initial
• support for hyphens between initials
• support for a separated surname 'prefix'
• looking for bad 'et al' placement (before an author group rather than after), which causes the author group to be ignored
• arXiv suppression: removing 'arxiv' or 'e-print arxiv' before a report number and after a title
• ed. ==> eds?.
• corrected placement of 'et al' and the last 'ed.' pattern (reversed)
• support for author groups in brackets
• fixes the problem with multiple report number splitting in a single reference line (introduced in ff97c0942af09d9298b70f0934582ac05a75dc65)