Recognize when libc++ has an implementation of parallel algorithms

[akukanov]

Currently, the oneDPL configuration macro that defines presence of C++17 execution policies in the standard library completely ignores libc++:

oneDPL/include/oneapi/dpl/pstl/onedpl_config.h

Lines 152 to 154 in 6cdcd54

	// Should be defined to 1 for environments with a vendor implementation of C++17 execution policies
	#define _ONEDPL_CPP17_EXECUTION_POLICIES_PRESENT \
	(_ONEDPL___cplusplus >= 201703L && (_MSC_VER >= 1912 || (_GLIBCXX_RELEASE >= 9 && __GLIBCXX__ >= 20190503)))

However, since LLVM 17 there is partial support for std parallelism in libc++, guarded by #if !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17. It's off by default but can be enabled as an experimental functionality by a special macro or a compiler flag (see https://libcxx.llvm.org/UserDocumentation.html#enabling-experimental-c-library-features).

So these conditions need to be added to the definition of our configuration macro.