Validate allocated DataNode is not null to prevent gcc -Wnull-dereference (issue #830)

[brawner]

gcc appears to warn about not checking for null when allocating these pointers. It is probably a false positive because clang does not issue the same warning, and optimizations can change how gcc issues this warning. Nonetheless, this is one approach to resolving the issue.

Addresses #830

[semanticdiff-com]

Review changes with SemanticDiff.

[jwellbelove]

There are errors in the CI
Example

/home/runner/work/etl/etl/test/../include/etl/map.h:1226:32: error: comparison of distinct pointer types ('etl::map_base::Node *' and 'etl::imap<std::basic_string<char>, int, std::greater<std::basic_string<char>>>::Data_Node **')
      inserted = inserted_node == &node;

[jwellbelove]

I've investigated this some more and found a pattern to the warning.
Many ETL containers use a function called allocate_data_node() to allocate a node from an etl::ipool. Only in the set/multiset/map/multimap containers does it return a reference; in all the others it returns a pointer. In all of the containers the function that calls it returns a reference (they don't seem to exhibit the warning), so the solution appears to be to just modify the allocate_data_node() in set/multiset/map/multimap. I've already applied this change to my issue branch and run the CI test set with -O2 optimisation.

[brawner]

@jwellbelove Did you want me to fix this PR, or are you handling it on your branch?

[jwellbelove]

I've tested my changes with my CI scripts here, so I'll merge them. Thanks for your PR though. All inputs are always appreciated.

[brawner]

I can confirm that your fixed version appears to work with my code as well.