Cleanup guiglobals

[lenhard]

Implements #1408

Cleans up the GUIGlobals class and move LatexFieldFormatter to the bibtex package, which is now independent of the gui package.

As a final step, I would move the class net.sf.jabref.importer.fileformat.FieldContentParser to the bibtex package, which would make the package self-contained (apart from imports to the preferences and related classes) and would move the whole bibtex package into logic. Are there objections to this?

[lenhard]

I restructured variable positioning as discussed in #1408 and turn MAX_BACK_HISTORY_SIZE, LINE_LENGTH, and INDENT into proper preferences.

@JabRef/developers What about turning the bibtex package into a subpackage of logic? Or should that rather stay a top-level package?

[simonharrer]

The more stuff is in the logic package, the better.

[lenhard]

Ok, from side this is good to go. I even was able to move some more classes into the model package.

[matthiasgeiger]

LGTM

[simonharrer]

This comment is placed wrongly, I think.

[tobiasdiez]

LGTM, just some minor remarks.

What I do not understand is the purpose of the bibtex package. Some of the classes seem like they do not have anything to do with writing or parsing bibtex and are rather general, for example the comparators. Maybe it would make sense to completely dissolve the bibtex package by moving the parsing logic to the import package and writing to export...in the end the bibtex format is just a special import-export format.

[lenhard]

@tobiasdiez Thanks for your remarks, I will integrate them as well.

As for the bibtex package: Its existence is a historical thing and I do not know the exact motivation behind. It seems like "bundle parsing and writing of pure bibtex", but, with the addition of new importers and exporters, responsibilities shifted and mixed over time. The big advantage we have now is that the package moved to logic, which means that there is a guarantee that it is independent of GUI code and it can be re-used for new GUI stuff with relative safety. If we move it out of logic (to importers or exporters) we loose that advantage. Ultimately, the goal should be to get the GUI independent code from the import/export package into logic as well.

Out of topic: I just noticed the codecov report. How the hell did code coverage increase by a wooping 11.27% with this PR? I didn't even write new tests. This has to be an error.

[lenhard]

I implemented the above comments and the PR is once again ready to go from my side.

One remark: COL_DEFINITION_FIELD_SEPARATOR cannot go to GUIGlobals, since this would create dependencies from logic to gui. Therefore, I moved it into COL_DEFINITION_FIELD_SEPARATOR to Globals. This is fair, I think, since the variables relates to our BibTeX sytnax and field content, which is independent of the GUI.

I still don`t get how the coverage can have increased that much.