Several changes for timestamp bundle:

Fully defined the structure.
Single copy of it + single timer across all modules, managing it is now the responsibility of KernelState.

add global_critical_region::PrepareToAcquire, which uses Prefetchw on the global crit. We now know we can use Prefetchw on all cpus that have AVX.
add  KeQueryInterruptTime, which is used by some dashboards.

add threading::NanoSleep

src/xenia/base/mutex.h

@@ -11,7 +11,7 @@
 #define XENIA_BASE_MUTEX_H_
 #include <mutex>
 #include "platform.h"
-
+#include "memory.h"
 #define XE_ENABLE_FAST_WIN32_MUTEX 1
 namespace xe {
 
@@ -149,6 +149,12 @@ class global_critical_region {
     return global_unique_lock_type(mutex());
   }
 
+  static inline void PrepareToAcquire() {
+#if XE_PLATFORM_WIN32 == 1
+    swcache::PrefetchW(&mutex());
+#endif
+  }
+
   // Acquires a deferred lock on the global critical section.
   static inline global_unique_lock_type AcquireDeferred() {
     return global_unique_lock_type(mutex(), std::defer_lock);