proper error handling

[Robert Jäschke]

This is a subtask of issue #2836.

Since the PDF extraction is quite complex, errors/exceptions can arise at different levels:

1. Grobit itself in the extraction step.
2. During the model conversion. (Which is currently implemented using the BibTeX parser.) E.g., due to missing or incompatible fields.
3. During enrichment using the BibSonomy database.

Errors need to be collected for each post and presented to the user. See how it is currently done in the BibTeXParser, when a user posts BibTeX that contains several entries.