[clang] [clang-repl] Delegate CodeGen related operations for PTU to IncrementalParser (PR #137458)

[Anutosh Bhat via cfe-commits]

anutosh491 wrote:

This works as expected now !

```
anutosh491 at vv-nuc:/build/anutosh491/llvm-project/build2/bin$ ./clang-repl --cuda
clang-repl> extern "C" int printf(const char*, ...);
clang-repl> template <typename T> __device__ inline T sum(T a, T b) { return a + b; }
clang-repl> __global__ void test_kernel(int* value) { *value = sum(40, 2); }
clang-repl> int var;
clang-repl> int* devptr = nullptr;
clang-repl> printf("cudaMalloc: %d\n", cudaMalloc((void **) &devptr, sizeof(int)));
cudaMalloc: 0
clang-repl> test_kernel<<<1,1>>>(devptr);
clang-repl> printf("CUDA Error: %d\n", cudaGetLastError());
CUDA Error: 0
clang-repl> printf("cudaMemcpy: %d\n", cudaMemcpy(&var, devptr, sizeof(int), cudaMemcpyDeviceToHost));
cudaMemcpy: 0
clang-repl> printf("Value: %d\n", var);
Value: 42
clang-repl> 
```

https://github.com/llvm/llvm-project/pull/137458