Subtract stream.start when getting the startXRef property for documents with a Linearization dictionary (issue 11330)

For documents with a Linearization dictionary the computed `startXRef` position will be relative to the raw file, rather than the actual PDF document itself (which begins with `%PDF-`).
Hence it's necessary to subtract `stream.start` in this case, since otherwise the `XRef.readXRef` method will increment the position too far resulting in parsing errors.

src/core/document.js

@@ -453,7 +453,7 @@ class PDFDocument {
       // Find the end of the first object.
       stream.reset();
       if (find(stream, 'endobj', 1024)) {
-        startXRef = stream.pos + 6;
+        startXRef = (stream.pos + 6) - stream.start;
       }
     } else {
       // Find `startxref` by checking backwards from the end of the file.