Tweak inlining heuristics.

go/lem driven:

Performance:
Richards +41%
CaffeineMethod +43%
ReversiBench: +52%
Towers: +73%
Tak: +85%

Memory use: 7% less memory
CompileTime: 14% increase
CodeSize: 8% increase

Last three measures are now more acceptable given we JIT.

Change-Id: Ic4aa6535d2b76cf3545ef00e9b2ae32330f10745
diff --git a/compiler/driver/compiler_options.h b/compiler/driver/compiler_options.h
index e6acab4..38c2927 100644
--- a/compiler/driver/compiler_options.h
+++ b/compiler/driver/compiler_options.h
@@ -52,7 +52,7 @@
   static const bool kDefaultGenerateDebugInfo = kIsDebugBuild;
   static const bool kDefaultIncludePatchInformation = false;
   static const size_t kDefaultInlineDepthLimit = 3;
-  static const size_t kDefaultInlineMaxCodeUnits = 20;
+  static const size_t kDefaultInlineMaxCodeUnits = 32;
   static constexpr size_t kUnsetInlineDepthLimit = -1;
   static constexpr size_t kUnsetInlineMaxCodeUnits = -1;