<xutility>: copy_backward() and move_backward() shouldn't have been parallelized

[StephanTLavavej]

We implemented:

STL/stl/inc/xutility

Lines 4260 to 4267 in 2787718

	#if _HAS_CXX17
	template <class _ExPo, class _BidIt1, class _BidIt2, _Enable_if_execution_policy_t<_ExPo> = 0>
	_BidIt2 copy_backward(_ExPo&&, _BidIt1 _First, _BidIt1 _Last, _BidIt2 _Dest) noexcept /* terminates */ {
	// copy [_First, _Last) backwards to [..., _Dest)
	// not parallelized as benchmarks show it isn't worth it
	return _STD copy_backward(_First, _Last, _Dest);
	}
	#endif // _HAS_CXX17

STL/stl/inc/xutility

Lines 4342 to 4349 in 2787718

	#if _HAS_CXX17
	template <class _ExPo, class _BidIt1, class _BidIt2, _Enable_if_execution_policy_t<_ExPo> = 0>
	_BidIt2 move_backward(_ExPo&&, _BidIt1 _First, _BidIt1 _Last, _BidIt2 _Dest) noexcept /* terminates */ {
	// move [_First, _Last) backwards to [..., _Dest)
	// not parallelized as benchmarks show it isn't worth it
	return _STD move_backward(_First, _Last, _Dest);
	}
	#endif // _HAS_CXX17

I don't see these overloads declared or defined in WG21-N4910, and I don't see them mentioned in the original paper P0024R2 anywhere (including in the table of algorithms being parallelized; notably, that table did mention the algorithms we missed in #525).

I think we went too far, and that we should remove these overloads and any associated test coverage. I am skipping them for Standard Library Modules as they are non-Standard.

[StephanTLavavej]

I'll create a PR.