Add method to automatically determine content type based on HTTP headers

This avoids parsing PDF as HTML in suche_st harvester. (At least after we deleted the three offending documents from the cache once.)