refextract: improve realign numeration

• Improve the realigning of numeration, across badly split affiliation lines (added config variable to specify an acceptable numeric gap between numerated affiliations, in the event of bad pdftotext conv.).

• Remove -raw from pdftotext conversion, to do: use -layout instead.

• Strip reference section from document when looking for auths/affs.

• partially able to obtain standard authors through comma and numeration placement