win32: Properly wait unil the thread finishes in dll unload

Without this we run into a problem. After 50ms we forcably kill
the pool thread. However, if the pool thread currently holds the
allocation lock, but doesn't get scheduled within 50ms, said lock
will never be released causing an infinite deadlock and preventing
the DLL from releasing. Further, in the unload scenario (as opposed
to process exit), we may actually want to let the thread pool finish
in case it's working on finishing an operation on another thread
(since the user may be expecting the answer).

driver/others/blas_server_win32.c

@@ -462,14 +462,7 @@ int BLASFUNC(blas_thread_shutdown)(void){
 
     for(i = 0; i < blas_num_threads - 1; i++){
       // Could also just use WaitForMultipleObjects
-      DWORD wait_thread_value = WaitForSingleObject(blas_threads[i], 50);
-
-#ifndef OS_WINDOWSSTORE
-      // TerminateThread is only available with WINAPI_DESKTOP and WINAPI_SYSTEM not WINAPI_APP in UWP
-      if (WAIT_OBJECT_0 != wait_thread_value) {
-        TerminateThread(blas_threads[i],0);
-      }
-#endif
+      DWORD wait_thread_value = WaitForSingleObject(blas_threads[i], INFINITE);
 
       CloseHandle(blas_threads[i]);
     }