[llvm] [DWARFLinker] Use different addresses to distinguish invalid DW_AT_LLVM_stmt_sequence offset (PR #149376)

[Peter Rong via llvm-commits]

DataCorrupted wrote:

The whole debugging process would be _much_ easier if I have a min repro when either is generated. So no at the moment. By the time I can add a test, I would be close to resolving this :\

https://github.com/llvm/llvm-project/pull/149376