inline_string reimplementation

- change implementation to template
- move functions definitions outside class

remove doubled clang-format configuration's rule

minor comments updates

Add new cases to fully cover inline_string

fix issues from klocwork and coverity

tests: add libunwind suppresion

Radix tree optimizations

Run docs update as separate job on GHA

Memcheck reports an error in libunwind exception raising.

The only difference between lower and upper bound is behavior in
case the element already exists.

to increase lower_bound coverage.

to helper methods.

Change bottom_leaf to return any leaf placed lower than a certain depth.
This way, leaf variable in internal_emplace will always point to an
element with specified key (if only it exists). It saves going descending
the tree for the second time when the element is an embedded entry.

Implementation of basic_string_view class

run most of CI jobs on GHA only

Parametrized version of original string_view

add remaining methods for radix_tree

In case there is already such element we can return it
just after descend - no need to descend the tree twice.

Radix tree improvements

radix_tree: refactor tagged_node_ptr to use self_relative_ptr

Previous type was copied from vmemcache and is too small.

std::atomic specialization for self_relative_pointer 

After changing radix_tree iterators API to return reference to the value
instead of a wrapper user should be able to check the capacity of an
inline string.

The storage cost of one additional field is not very high and it also
allows for checks inside assign() methods of inline_string.

and enable swap tests for radix_tree