[BOLT][NFC] Fix data race in ShrinkWrapping stats

[Amir]

Summary

Fix data race reported by ThreadSanitizer in clang.test:

ThreadSanitizer: data race /data/llvm-project/bolt/lib/Passes/ShrinkWrapping.cpp:1359:28
in llvm::bolt::ShrinkWrapping::moveSaveRestores()

The issue is with incrementing global counters from multiple threads.

bolt/include/bolt/Passes/ShrinkWrapping.h

class ShrinkWrapping {
   std::vector<int64_t> PopOffsetByReg;
   std::vector<MCPhysReg> DomOrder;
   CalleeSavedAnalysis CSA;
   std::vector<SmallSetVector<MCInst *, 4>> SavePos;
   std::vector<uint64_t> BestSaveCount;
   std::vector<MCInst *> BestSavePos;
 
   /// Pass stats
-  static uint64_t SpillsMovedRegularMode;
-  static uint64_t SpillsMovedPushPopMode;
+  static std::atomic_uint64_t SpillsMovedRegularMode;
+  static std::atomic_uint64_t SpillsMovedPushPopMode;
 
   Optional<unsigned> AnnotationIndex;
 
   /// Allow our custom worklist-sensitive analysis
   /// PredictiveStackPointerTracking to access WorklistItem
 public:
   struct WorklistItem {
     enum ActionType : uint8_t {

bolt/lib/Passes/ShrinkWrapping.cpp

 }
 
 void StackLayoutModifier::initialize() {
   classifyStackAccesses();
   classifyCFIs();
   IsInitialized = true;
 }
 
-uint64_t ShrinkWrapping::SpillsMovedRegularMode = 0;
-uint64_t ShrinkWrapping::SpillsMovedPushPopMode = 0;
+std::atomic_uint64_t ShrinkWrapping::SpillsMovedRegularMode{0};
+std::atomic_uint64_t ShrinkWrapping::SpillsMovedPushPopMode{0};
 
 using BBIterTy = BinaryBasicBlock::iterator;
 
 void ShrinkWrapping::classifyCSRUses() {
   DominatorAnalysis<false> &DA = Info.getDominatorAnalysis();
   StackPointerTracking &SPT = Info.getStackPointerTracking();
   UsesByReg = std::vector<BitVector>(BC.MRI->getNumRegs(),
                                      BitVector(DA.NumInstrs, false));