[MLIR] Add extra locking during cubin generation.

[herhut]

Summary

We also need to lock the LLVMDialect mutex when initializing
LLVM targets or destroying llvm modules concurrently. Added another
scoped lock to that effect.

mlir/lib/Conversion/GPUToCUDA/ConvertKernelFuncToCubin.cpp

 #include "mlir/Target/NVVMIR.h"
 
 #include "llvm/ADT/Optional.h"
 #include "llvm/ADT/Twine.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/LegacyPassManager.h"
 #include "llvm/IR/Module.h"
 #include "llvm/Support/Error.h"
+#include "llvm/Support/Mutex.h"
 #include "llvm/Support/TargetRegistry.h"
 #include "llvm/Support/TargetSelect.h"
 #include "llvm/Target/TargetMachine.h"
 
 using namespace mlir;
 
 namespace {
 // TODO(herhut): Move to shared location.
: public PassWrapper<GpuKernelToCubinPass,
                          OperationPass<gpu::GPUModuleOp>> {
 public:
   GpuKernelToCubinPass(CubinGenerator cubinGenerator)
       : cubinGenerator(cubinGenerator) {}
 
   void runOnOperation() override {
     gpu::GPUModuleOp module = getOperation();
 
+    // Lock access to the llvm context.
+    llvm::sys::SmartScopedLock<true> scopedLock(
+        module.getContext()
+            ->getRegisteredDialect<LLVM::LLVMDialect>()
+            ->getLLVMContextMutex());
+
     // Make sure the NVPTX target is initialized.
     LLVMInitializeNVPTXTarget();
     LLVMInitializeNVPTXTargetInfo();
     LLVMInitializeNVPTXTargetMC();
     LLVMInitializeNVPTXAsmPrinter();
 
     auto llvmModule = translateModuleToNVVMIR(module);
     if (!llvmModule)