Implement irreducible loop support in optimizing.

So we don't fallback to the interpreter in the presence of
irreducible loops.

Implications:
- A loop pre-header does not necessarily dominate a loop header.
- Non-constant redundant phis will be kept in loop headers, to
  satisfy our linear scan register allocation algorithm.
- while-graph optimizations, such as gvn, licm, lse, and dce
  need to know when they are dealing with irreducible loops.

Change-Id: I2cea8934ce0b40162d215353497c7f77d6c9137e
diff --git a/compiler/optimizing/inliner.cc b/compiler/optimizing/inliner.cc
index 48d3299..293282e 100644
--- a/compiler/optimizing/inliner.cc
+++ b/compiler/optimizing/inliner.cc
@@ -539,7 +539,7 @@
     return false;
   }
 
-  if (callee_graph->TryBuildingSsa(handles_) != kBuildSsaSuccess) {
+  if (callee_graph->TryBuildingSsa(handles_) != kAnalysisSuccess) {
     VLOG(compiler) << "Method " << PrettyMethod(method_index, callee_dex_file)
                    << " could not be transformed to SSA";
     return false;