[turbofan] Instruction scheduler: keep ready nodes with same latency sorted in original order.

src/compiler/instruction-scheduler.cc

Index: src/compiler/instruction-scheduler.cc
diff --git a/src/compiler/instruction-scheduler.cc b/src/compiler/instruction-scheduler.cc
index d5bd1255767ff9351924e08ccf1c4566a172284a..a49e3a557d09ef80b0adb692660b47cd17b320e6 100644
--- a/src/compiler/instruction-scheduler.cc
+++ b/src/compiler/instruction-scheduler.cc
@@ -17,7 +17,7 @@ void InstructionScheduler::SchedulingQueueBase::AddNode(
   // the next best candidate to schedule.
   auto it = nodes_.begin();
   while ((it != nodes_.end()) &&
-         ((*it)->total_latency() > node->total_latency())) {
+         ((*it)->total_latency() >= node->total_latency())) {
     ++it;
   }
   nodes_.insert(it, node);