Compile-time tuning

Specialized the dataflow iterators and did a few other minor tweaks.
Showing ~5% compile-time improvement in a single-threaded environment;
less in multi-threaded (presumably because we're blocked by something
else).

Change-Id: I2e2ed58d881414b9fc97e04cd0623e188259afd2
diff --git a/compiler/dex/vreg_analysis.cc b/compiler/dex/vreg_analysis.cc
index 07f37bb..2551068 100644
--- a/compiler/dex/vreg_analysis.cc
+++ b/compiler/dex/vreg_analysis.cc
@@ -444,7 +444,7 @@
   }
 
   /* Do type & size inference pass */
-  PreOrderDfsIterator iter(this, true /* iterative */);
+  RepeatingPreOrderDfsIterator iter(this);
   bool change = false;
   for (BasicBlock* bb = iter.Next(false); bb != NULL; bb = iter.Next(change)) {
     change = InferTypeAndSize(bb);