Leak FP when using a lambda capture

https://godbolt.org/z/W3Y847sMT
```c++
struct Ptr {
int* p;
explicit Ptr(int* p) : p(p) {}
Ptr(Ptr&& o) : p(o.p) { o.p = nullptr; }
~Ptr() { delete p; }
};
void top() {
auto l = [q = Ptr(new int)] {}; // FP: Potential leak of memory pointed to by field 'p' [cplusplus.NewDeleteLeaks]
l();
}
```

A similar issue was in the past #60896 for unique_ptr leaks and those were suppressed in #152751 but that only suppresses for standard smart-pointers, thus a custom uptr like class like `Ptr` here still warns.

Start by reproducing the Godbolt example and inspect the cplusplus.NewDeleteLeaks checker behavior for the lambda capture. Compare the handling with the suppression added for #60896 in #152751. Done means the custom Ptr example no longer produces this false-positive leak warning, with a regression test covering it.

Written by the indexing model from the issue text.