[CIR][ABI][AArch64][Lowering] Fix the callsite for nested unions

[bruteforceboy]

For example, the following reaches "NYI" when lowering to AArch64:

typedef struct {
  union {
    struct {
      char a, b;
    };
    char c;
  };
} A;

void foo(A a) {}

void bar() {
  A a;
  foo(a);
}

Currently, the value of the struct becomes a bitcast operation, so we can simply extend findAlloca to be able to trace the source alloca properly, then use that for the coercion through memory. I have also added a test for this case.

[smeenai]

Looks like these are dead, at least in the IR parts you're checking? (It might be useful to use CHECK-NEXT to indicate when you're checking the whole IR vs. portions of it.) Not a big deal, optimizations should be able to easily remove them; I guess it's just a result of the extra type casts we need to do at the CIR level.

[bcardosolopes]

True! Maybe if it's not relevant just use {{.*}} instead of giving it a name next time!

[bruteforceboy]

noted