parented_ptr: Assert parent at destruction instead of construction

[haslersn]

Previously, parented_ptr asserted on construction that the pointed-to object has a parent. However, sometimes one wants to create an object and assign it to a parent later. The recommendation was to create a unique_ptr p instead, later assign the object to a parent and then create a new variable using make_parented(p.release()).

This has however several drawbacks:

• There is a short period where the object is already assigned to a parent but still owned by a unique_ptr. This can lead to a double free.
• There are two pointers that and it depends on the code location which one is the correct one to access the object.
• It is inconvenient to use, because one has to write an extra line where the unique_ptr is released.

With this commit, parented_ptr asserts that the pointed-to object has a parent at the destruction of the parented_ptr. That means, that one can store the object using a parented_ptr a priori and there will be a warning (from the failing DEBUG_ASSERT) in case any object is leaked.

Optionally, in a future PR, this behaviour can even be changed so as to delete the object that is going to be leaked. That is, the parented_ptr destructor wouldn't only DEBUG_ASSERT that the pointed-to object has a parent, but it would also delete it otherwise.

Additional changes:

• Made parented_ptr implicitly convertible to a raw pointer. The reasons why the standard library types are not convertible to a raw pointer don't apply to parented_ptr. Regarding the insufficient usage of parented_ptr in Mixxx at the moment, making it implicitly convertible will make its adoption easier because less code has to be changed. (No .get() everywhere.)
• Added a constructor taking nullptr_t.
• Added noexcept where applicable.
• Formatted the code using clang-format.

[uklotzde]

@haslersn Would you please sign our contributor agreement that allows us to publish your code:
https://docs.google.com/forms/d/e/1FAIpQLScC9QG327sjLL0eWftmfGUasxFFLxg6LCXJ2xHDYRzFIRqyiw/viewform?formkey=dEpYN2NkVEFnWWQzbkFfM0ZYYUZ5X2c6MQ

I confused you with another first-time contributor who provided the NixOS files a few days ago.

[uklotzde]

Good catch, I didn't notice either.

[uklotzde]

LGTM. Thank you for improving the overall code quality and for digging deep to fix edge cases!

[Be-ing]

I'm getting build errors now with Clang 8.0.0 on Fedora 30:

src/util/parented_ptr.h:23:28: error: expected ';' at end of declaration list
    parented_ptr(nullptr_t) noexcept
                           ^
                           ;

...

src/util/parented_ptr.h:60:29: error: unknown type name 'nullptr_t'; did you mean 'std::nullptr_t'?
    parented_ptr& operator=(nullptr_t) noexcept {
                            ^~~~~~~~~
                            std::nullptr_t
/usr/bin/../lib/gcc/x86_64-redhat-linux/9/../../../../include/c++/9/x86_64-redhat-linux/bits/c++config.h:2328:29: note: 
      'std::nullptr_t' declared here
  typedef decltype(nullptr)     nullptr_t;
                                ^
In file included from src/main.cpp:26:
In file included from src/mixxx.h:30:
In file included from src/util/timer.h:9:
src/util/parented_ptr.h:23:18: error: field has incomplete type 'parented_ptr<ControlProxy>'
    parented_ptr(nullptr_t) noexcept
                 ^
src/util/timer.h:132:32: note: in instantiation of template class 'parented_ptr<ControlProxy>' requested here
    parented_ptr<ControlProxy> m_pGuiTick;
                               ^
src/util/parented_ptr.h:17:7: note: definition of 'parented_ptr<ControlProxy>' is not complete until the closing '}'
class parented_ptr {
      ^