[CIR] Lowering of no-proto aliases/functions not right?

See this patch here: https://github.com/llvm/llvm-project/pull/212308

Particularly the attr-alias-no-proto.c test. We end up having a difference in the LLVM IR generated that we should probably fix. I suspect that our lowering of functions/aliases/calls isn't properly setting the 'variadic-ness' of these functions properly.

Start with the attr-alias-no-proto.c test and compare the generated LLVM IR with the patch at https://github.com/llvm/llvm-project/pull/212308. Trace the lowering of the functions, aliases, and calls involved, focusing on whether their variadic status is set consistently. Done means the IR difference is resolved and the test passes with the expected output.

Written by the indexing model from the issue text.