Make the HTMLDocumentParser yield more aggressively

Source/core/html/parser/HTMLParserScheduler.cpp

Index: Source/core/html/parser/HTMLParserScheduler.cpp
diff --git a/Source/core/html/parser/HTMLParserScheduler.cpp b/Source/core/html/parser/HTMLParserScheduler.cpp
index 61cb172ccae8990de673cc9d4634963ad5e2d19c..0765052e0a77a31cc68ab312dfe6a5cb52f4f2be 100644
--- a/Source/core/html/parser/HTMLParserScheduler.cpp
+++ b/Source/core/html/parser/HTMLParserScheduler.cpp
@@ -93,6 +93,12 @@ void HTMLParserScheduler::continueNextChunkTimerFired(Timer<HTMLParserScheduler>
 
 void HTMLParserScheduler::scheduleForResume()
 {
+    // FIXME: Currently Timers use the SharedTimer queue, so scheduling
+    // a 0ms timer here may cause the parser to resume again before
+    // returning to the Chromium MessageLoop, thus starving other tasks
+    // in the MessageLoop. This will be fixed by moving Blink's Timer
+    // class ontop of postTask, but until then we're probably missing
+    // frames with this implementation.
     m_continueNextChunkTimer.startOneShot(0, FROM_HERE);
 }