tracker-extract: Process small / empty text files

Previously small or empty text files were ignored by the extractor. This could cause inconsistent results in the database, for example if the user truncated a regular text file to 0 bytes then the old content would remain in the store.

We now always update the nie:plainTextContent field for small and 0-byte files.

In 3.0 we should fix #78 (closed) so this shouldn't have much performance impact. Perhaps we shouldn't backport it to 2.3, as nobody has reported a specific issue and it could cause performance regressions there.