Disable range requests if the origin ignores byte-range.

Use the non-linearized PDF code path if the server doesn't respond to range
requests with a partial response. Previously, Chrome entered a request loop if
the PDF was linearized and the server advertized byte range support but
responded with a 200 code instead of 206.

BUG=587502
R=thestig@chromium.org
TEST=Load https://www.usenix.org/legacy/publications/library/proceedings/sec96/full_papers/goldberg/goldberg.pdf

Review URL: https://codereview.chromium.org/1705353002

Cr-Commit-Position: refs/heads/master@{#376053}

pdf/document_loader.cc

@@ -383,7 +383,7 @@ void DocumentLoader::DidOpen(int32_t result) {
     multipart_boundary_ = boundary;
   } else {
     // Need to make sure that the server returned a byte-range, since it's
-    // possible for a server to just ignore our bye-range request and just
+    // possible for a server to just ignore our byte-range request and just
     // return the entire document even if it supports byte-range requests.
     // i.e. sniff response to
     // http://www.act.org/compass/sample/pdf/geometry.pdf
@@ -393,6 +393,8 @@ void DocumentLoader::DidOpen(int32_t result) {
       current_pos_ = start_pos;
       if (end_pos && end_pos > start_pos)
         current_chunk_size_ = end_pos - start_pos + 1;
+    } else {
+      partial_document_ = false;
     }
   }