[clang] [Sema] Mark alias/ifunc targets used and consider mangled names (PR #87130)

[Fangrui Song via cfe-commits]

MaskRay wrote:

I really appreciate the suggestions. `alias-unused.cpp` and `alias-unused-win.cpp` contain test improvement that should be pre-commited once they look good enough.
Then this PR can be changed to show the difference.

On a separate note, I wanted to clarify that `-Wunused-function` false positives/negatives shouldn't be automatically considered security bugs.
Categorizing all these warning improvements as "security bugs" would dilute the meaning of "security bugs" and make it harder to prioritize real vulnerabilities.

(
```
What is a GCC security bug?
===========================

    A security bug is one that threatens the security of a system or
    network, or might compromise the security of data stored on it.
    In the context of GCC, there are multiple ways in which this might
    happen and some common scenarios are detailed below.
```
)

(An attacker can easily bypass warnings: remove `-Werror` (uncommon in distros anyway), remove `-Wall` (which covers `-Wunused-function`, or use a pragma to disable `-Wunused-function` locally.
)

The description contains an example about name mangling differences (https://github.com/llvm/llvm-project/pull/87130/files#r1554029811)
and I mentioned that "This inconsistency makes alias/ifunc difficult to use in C++ with portability."

```
extern "C" {
static void f0() {}
// GCC: void g0() __attribute__((alias("_ZL2f0v")));
// Clang: void g0() __attribute__((alias("f0")));
}
```

I added microsoftDemangle tests to show the current behavior.
Since the feature that demangles to the function name without parameters (f3 instead of f3(int)) appears to be missing, I cannot address -Wunused-function false positives for microsoftDemangle with reasonable time complexity.


https://github.com/llvm/llvm-project/pull/87130