Empty `dc:title` metadata field in PDFs should be ignored

[judell]

I created an annotation on http://awspntest.apa.org/fulltext/2018-18843-001.pdf. It shows up untitled in activity pages:

Here's why: http://jonudell.net/h/apa-untitled-document.mp4

It happens here. We acquire the HTML doctitle as expected, but then overwrite it because dc:title exists in HTML metadata, even though it's empty.

[robertknight]

I reworded this to clarify that it relates to PDFs, not to HTML documents.