[clang] cuda clang: Move nvptx-surface.cu test to CodeGenCUDA (PR #134758)

[Austin Schuh via cfe-commits]

================
@@ -2,6 +2,170 @@
 // RUN: %clang_cc1 -triple nvptx64-unknown-unknown -fcuda-is-device -O3 -o - %s -emit-llvm | FileCheck %s
 #include "Inputs/cuda.h"
 
+struct char1 {
----------------
AustinSchuh wrote:

I tried that round 1 of this review, and it went quite poorly:  https://buildkite.com/llvm-project/github-pull-requests/builds/166089  I was not fully sure what the existing tests were testing well enough to feel comfortable changing them, and why this would make them fail.  I can keep poking, but would love any hints if you see anything.

https://github.com/llvm/llvm-project/pull/134758