Bug 1378537 - Store PerformanceEntry objects in AutoTArray; r=smaug

Storing PerformanceEntry objects in an AutoTArray (currently sized to
match our limit for resource timing entries) should omit some
per-entry allocations and give us a small speedup with creating
entries in tight loops.

MozReview-Commit-ID: LNgVhMn461g

--HG--
extra : rebase_source : 10250c2fad5ebaf0e89bf9759d157b7d1d855ffa

dom/performance/Performance.h

@@ -152,11 +152,14 @@ protected:
   nsTObserverArray<PerformanceObserver*> mObservers;
 
 private:
-  nsTArray<RefPtr<PerformanceEntry>> mUserEntries;
-  nsTArray<RefPtr<PerformanceEntry>> mResourceEntries;
+  static const uint64_t kDefaultResourceTimingBufferSize = 150;
+
+  // When kDefaultResourceTimingBufferSize is increased or removed, these should
+  // be changed to use SegmentedVector
+  AutoTArray<RefPtr<PerformanceEntry>, kDefaultResourceTimingBufferSize> mUserEntries;
+  AutoTArray<RefPtr<PerformanceEntry>, kDefaultResourceTimingBufferSize> mResourceEntries;
 
   uint64_t mResourceTimingBufferSize;
-  static const uint64_t kDefaultResourceTimingBufferSize = 150;
   bool mPendingNotificationObserversTask;
 
   RefPtr<PerformanceService> mPerformanceService;