Add better error handling

[ajenhl]

Currently there are exceptions that are not handled (MemoryError with database operations, FileExistsError with corpus stripping) and should be. Also, unanticipated errors should be caught and handled generically (log exception name and message, at least).

[ajenhl]

When a database is missing entries for texts in the supplied corpus (as referenced in the supplied catalogue), a warning log message is emitted, but it should really raise an exception, since that is likely to happen later (unless the text is unlabelled in the catalogue).