Improve embedded file name extraction in PDFParser

When we extract embedded files from PDFs, we are currently using the key in the PDEmbeddedFilesNameTreeNode as the file name that we store as the value of Metadata.RESOURCE_NAME_KEY in the embedded document's metadata.

I think we should try to get the file name from PDComplexFileSpecification's getFilename() first. If that is null, then we should fall back to the key value.