On my not-so-fast NUC it improves build time:

clean build without ccache (and first build with ccache):
real	6m15.062s
user	19m45.633s
sys	1m48.421s

second clean build with ccache:
real	0m48.836s
user	1m48.793s
sys	0m35.407s

It's not needed for CI, as builds are done in containers and do not share
ccache anyway

examples: update min cmake version to 3.4

fix default bytes_view in radix_tree

radix_tree improvements

use try_emplace instead of emplace

Merge changes from stable 1.11 to master

in examples using 'include' command - it also uses 'CheckIncludeFile'
which (in CMake <=3.3) worked only with set project language (CXX in this case),
and some included files are supposed to be checked with C lang, too.

With help of: https://github.com/nlohmann/cmake_min_version

CI: fix GHA docs' job - use branches names not tags

update Fedora image for doc job

Radix tree assing val fix

Fix value() function which was creating an empty string (wrong input parameter)
and add testcases for basic_inline_string<uint8_t> and basic_inline_string<wchar_t>
as value types.

Realocating path should be taken for values which are instances
of basic_inline_string<CharT, Traits> (not only for CharT = char).

add new release step to update version in vcpkg

readme: add info about minimal VS version for radix

coloring in IDE was not applied, because of this additional new line

docker: bump images' tag to 1.12

add stable-1.11 to valid branches

tests: string_view_assignment

cmake: look for libpmemobj also when building benchmarks

radix_tree: fix upper_bound implementation

- string_view_assignment.pass test port
- T_size_size.pass test port

In internal_bound we actually rely on leaf node to be the leftmost one.

If a key was not present in a radix tree and a slot variable was
set to &root, next_leaf function lead to accessing uninitialzied memory
(we tried to increment slot, as if it was a pointer to an element in array).