Macrobenchmark tweaks & gpu memory dumping

Add support for glob matching
Ensure glob matches are alphabetically sorted
Add feature to dump GPU memory usage after a test pass
Adjust gpu memory dump to be a bit more compact (skip empty outputs)

Test: this
Bug: 187718492
Change-Id: I6dc80b2d3379d8d10001116e1240727d9914bc10
diff --git a/libs/hwui/renderthread/RenderProxy.h b/libs/hwui/renderthread/RenderProxy.h
index 288f555..662b445 100644
--- a/libs/hwui/renderthread/RenderProxy.h
+++ b/libs/hwui/renderthread/RenderProxy.h
@@ -109,7 +109,7 @@
     // Not exported, only used for testing
     void resetProfileInfo();
     uint32_t frameTimePercentile(int p);
-    static void dumpGraphicsMemory(int fd);
+    static void dumpGraphicsMemory(int fd, bool includeProfileData = true);
 
     static void rotateProcessStatsBuffer();
     static void setProcessStatsBuffer(int fd);