Core dump on SCHED_OTHER

This was happening on the RT scheduler and causing significant problems
when a process would crash.

Change-Id: Idec897f43daaf835a601c8a77dfe7f9bfafd62da
diff --git a/third_party/google-glog/src/signalhandler.cc b/third_party/google-glog/src/signalhandler.cc
index c3a0f80..049efa5 100644
--- a/third_party/google-glog/src/signalhandler.cc
+++ b/third_party/google-glog/src/signalhandler.cc
@@ -308,6 +308,14 @@
       sleep(1);
     }
   }
+
+  {
+    // Put this back on SCHED_OTHER by default.
+    struct sched_param param;
+    param.sched_priority = 0;
+    sched_setscheduler(0, SCHED_OTHER, &param);
+  }
+
   // This is the first time we enter the signal handler.  We are going to
   // do some interesting stuff from here.
   // TODO(satorux): We might want to set timeout here using alarm(), but