SWDEV-352197 - Destroy virtual device in thread destructor

Windows kills threads on exit without any notification. However,
runtime can still destroy VirtualGPU object from the host thread with
HostQueue destruction.
This change also forces RGP trace transfer on the last capture without
any delays.

Change-Id: I768e87e99e1d23a021e63c12f36e450817743759


[ROCm/clr commit: ad33a021cb]

projects/clr/rocclr/device/pal/palvirtual.cpp

@@ -1072,15 +1072,15 @@ bool VirtualGPU::allocHsaQueueMem() {
 }
 
 VirtualGPU::~VirtualGPU() {
+  // Not safe to remove a queue. So lock the device
+  amd::ScopedLock k(dev().lockAsyncOps());
+  amd::ScopedLock lock(dev().vgpusAccess());
+
   // Destroy RGP trace
   if (rgpCaptureEna()) {
     dev().rgpCaptureMgr()->FinishRGPTrace(this, true);
   }
 
-  // Not safe to remove a queue. So lock the device
-  amd::ScopedLock k(dev().lockAsyncOps());
-  amd::ScopedLock lock(dev().vgpusAccess());
-
   while (!freeCbQueue_.empty()) {
     auto cb = freeCbQueue_.front();
     delete cb;