Remove cuptiFinalize.

In cupti samples, only cuptiFlush is used.
I can't find any places calling cuptiFinalize and
this API can error out as not_implemented in some
cuda installation.

paddle/fluid/platform/device_tracer.cc

@@ -322,7 +322,6 @@ class DeviceTracerImpl : public DeviceTracer {
     DisableActivity();
     dynload::cuptiUnsubscribe(subscriber_);
     CUPTI_CALL(dynload::cuptiGetTimestamp(&end_ns_));
-    PADDLE_ENFORCE(dynload::cuptiFinalize());
     enabled_ = false;
   }
 

paddle/fluid/platform/dynload/cupti.h

@@ -72,7 +72,6 @@ extern void *cupti_dso_handle;
   __macro(cuptiGetResultString);              \
   __macro(cuptiActivityGetNumDroppedRecords); \
   __macro(cuptiActivityFlushAll);             \
-  __macro(cuptiFinalize);                     \
   __macro(cuptiSubscribe);                    \
   __macro(cuptiUnsubscribe);                  \
   __macro(cuptiEnableCallback);               \