Also report peak malloc memory usage from V8

gin/v8_isolate_memory_dump_provider.cc

Index: gin/v8_isolate_memory_dump_provider.cc
diff --git a/gin/v8_isolate_memory_dump_provider.cc b/gin/v8_isolate_memory_dump_provider.cc
index 601937b45f0a2820b0789c14a90263016df9448b..d92f82b9e2052db8c433f34ce78c9fd46dcdb39d 100644
--- a/gin/v8_isolate_memory_dump_provider.cc
+++ b/gin/v8_isolate_memory_dump_provider.cc
@@ -172,6 +172,14 @@ void V8IsolateMemoryDumpProvider::DumpHeapStatistics(
                                           system_allocator_name);
   }
 
+  std::string peak_malloc_name = dump_base_name + "/malloc_peak";

[primiano CORP (USE chromium), 2016/07/18 10:58:00]

I think the best approach would be reporting this as a column not a row,
otherwise would create funky double-counting issues, where we count both the
current and peak value.
Concretely this means turning this CL into a one-line (% wrapping) as follows:

malloc_dump->AddScalar("peak_size",
base::trace_event::MemoryAllocatorDump::kUnitsBytes,
heap_statistics.peak_malloced_memory())

+  auto* peak_malloc_dump =
+      process_memory_dump->CreateAllocatorDump(peak_malloc_name);
+  peak_malloc_dump->AddScalar(
+      base::trace_event::MemoryAllocatorDump::kNameSize,
+      base::trace_event::MemoryAllocatorDump::kUnitsBytes,
+      heap_statistics.peak_malloced_memory());
+
   // Add an empty row for the heap_spaces. This is to keep the shape of the
   // dump stable, whether code stats are enabled or not.
   auto* heap_spaces_dump =