Several Tick counter improvements.

webrtc/modules/utility/source/process_thread_impl_unittest.cc

Index: webrtc/modules/utility/source/process_thread_impl_unittest.cc
diff --git a/webrtc/modules/utility/source/process_thread_impl_unittest.cc b/webrtc/modules/utility/source/process_thread_impl_unittest.cc
index 34ed5c5a3845ddbd5adc2939f2136e512d0105a2..e1edd6ad96fad3033d9e7e7cffd52bd21ad1180c 100644
--- a/webrtc/modules/utility/source/process_thread_impl_unittest.cc
+++ b/webrtc/modules/utility/source/process_thread_impl_unittest.cc
@@ -253,6 +253,12 @@ TEST(ProcessThreadImpl, WakeUp) {
   MockModule module;
   int64_t start_time = 0;

[pbos-webrtc, 2015/11/12 11:55:41]

Shouldn't you initialize these to -1 and update the asserts below instead (that
it's > -1)?

[noahric, 2015/11/12 18:10:15]

I think we should actually just the startTime != something check;
MillisecondTimestamp and family make no claims about the absolute values of the
value they return, so this could legally return anything for the first value,
including -1. That's what I exploited when writing this originally (deltas from
zero mean small values, so loss of precision is acceptable over certain
timeframes).

   int64_t called_time = 0;
+
+  // On Mac, timestamps start at 0 the first time TickTime::Foo is called.
+  // Let this increment a very small amount before comparing our elapsed times.
+  (void)TickTime::MillisecondTimestamp();
+  started->Wait(10);
+
   // Ask for a callback after 1000ms.
   // TimeUntilNextProcess will be called twice.
   // The first time we use it to get the thread into a waiting state.