Keyword separator behaviour (again, from a groups perspective)

[oscargus]

Let me rephrase this: the use of the keyword separator is ambiguous in JabRef.

It is by default set to ,. This leads to:

1. StringTokenizer splits on both , and
2. when adding keywords, is used as a separator

Clearly these two behaviors are just by luck sometimes compatible. Based on 1. it would be better to have , as the default.

In addition, the EntryUtil method for splitting on keywords used by BibEntry totally ignores this.

Relates to #705

( @Siedlerchr 's comment was for the previous wording of this, although it is still relevant.)

I can in this context note that quite a bit of the refactoring I've made recently has been to pass this separator down to the correct methods. Except for the fact that I know that it is bad, I would almost suggest having the separator as a global static variable in model...

[Siedlerchr]

String Tokenizer ist really really legacy code. Should either be String.split or scanner
http://stackoverflow.com/questions/691184/scanner-vs-stringtokenizer-vs-string-split

[oscargus]

And in MedlineImporter the author decided to use ; as separator and ignore the preferences (the separator from the preferences is available in ImportFormatPreferences).

[koppor]

Addressed in #1950