No dangling assignment involving lifetimebound argument

[usx95]

https://godbolt.org/z/oedcc359z

struct S { int *x; };

S foo(const int &x [[clang::lifetimebound]]);

void usage() {
    S y = foo(1);  // Warning. Good.
    y = foo(1);    // No warning. Bad.
}

[higher-performance]

I think I'm confused. Neither of those warn for me, but should they? They do warn if the return type is a reference or a pointer, but not if it's an integer.

[usx95]

Modified the example for a better example. Basically we should warn in all cases because "lifetimebound" means we return an entity which refers to the arg. It does not have to be that we return the same reference as we accept.

[higher-performance]

Yeah makes sense, I agree, thanks!

[higher-performance]

I think #112614 should fix this.

[usx95]

Copying relevant text from comment on PR:

We already do a good job when GSL pointers and lifetimebound interact as in the bug: https://godbolt.org/z/9M9PsEGs7

I feel we need to fix the analysis for assignment operator here which would fix cases beyond std::string_view.
cc: @Xazax-hun @hokein