Do not use document.title as a fallback for title in PDFs

If a PDF has no embedded title, don't use `document.title` as a
fallback. In top-level frames PDF.js sets the title to the filename in the URL,
so this worked out OK. When PDF.js was embedded in an iframe however it
did not modify `document.title` and it would typically be set to a
generic default such as "PDF.js viewer" (the default title of the PDF.js
viewer application). The one case where we might actually want to use
`document.title` is in a custom viewer where the publisher has taken
care to set it appropriately based on some other source. This is a tiny
minority of usage currently.

Fixes #3372

src/annotator/integrations/pdf-metadata.js

@@ -133,7 +133,7 @@ export class PDFMetadata {
    */
   getMetadata() {
     return this._loaded.then(app => {
-      let title = document.title;
+      let title = '';
 
       if (
         app.metadata &&

src/annotator/integrations/test/pdf-metadata-test.js

@@ -320,5 +320,18 @@ describe('PDFMetadata', function () {
 
       assert.equal(metadata.title, 'Some title');
     });
+
+    it('returns an empty string if the PDF has no title', async () => {
+      const { pdfMetadata } = createPDFMetadata({
+        url: 'https://example.com/test.pdf',
+      });
+
+      // Earlier versions of the client used `document.title` as a fallback,
+      // but we changed this. See https://github.com/hypothesis/client/issues/3372.
+      document.title = 'Ignore me';
+      const metadata = await pdfMetadata.getMetadata();
+
+      assert.equal(metadata.title, '');
+    });
   });
 });