Allows to e.g. convert `std::u8string`s, translate them and run boundary
analysis.
Formatting isn't possible due to lack of facets
(more specifically: their IDs) in the standard libraries
Clang 10 & 11 on CI trigger the warning in C++20 mode for e.g. `l < r` with both being
strings.
Similar for `sl/sr` in `TEST_EQ((sl <= sr), (val <= 0));`
Move the skip condition into `test_main` to avoid missing compile failures due to wrong/missing includes in the headers and to improve formatting due to the removed indentation level of the PP macros
ICU 62+ returns only a "number" classification for a word ending in a number (like "Windows7") where previous versions returned "word&number"
Also refactor the test a bit to make error analysis easier
Gets rid of the `test_` prefix in the filenames and avoids confusing those with the actual tests especially by renaming the test framework file from `test_locale.hpp` to `unit_test.hpp`
Use a common `main` function calling into custom `test_main` functions and
move all common code into the former.
Also introduce an RAII class for a `locale_t` to avoid leaks and reduce
the code further.
Add some LCOV annotations for lines not expected to be reached.
For consistency the following rules are used:
- public includes (include folder) either include any <boost/locale/*>
first. At least <boost/locale/config.hpp>
- The first include in a cpp file is the corresponding header if any
- Then <boost/locale/*>, <boost/*>, <*>, "*" in this order
- Warning suppression pragmas come last where possible
- Use the C++ headers where available (i.e. <cstring> not <string.h>)
