Various header cleanups

[StephanTLavavej]

• <deque>, <forward_list>, <list>: Use nullptr instead of _Mapptr() and _Nodeptr().
  • At first, I was going to change these empty parens to empty braces for consistency with our modern conventions (braces make it clear when we're constructing a temporary instead of calling a function), but nullptr is even clearer.
  • This is guaranteed to work because they're "nullable pointers", and we're already using nullptr with these types:
    

    STL/stl/inc/deque

    Line 771 in ea32e86

    _Right_data._Map = nullptr;

    
    

    STL/stl/inc/forward_list

    Line 841 in ea32e86

    _Ptr->_Next = nullptr;

    
    

    STL/stl/inc/list

    Line 445 in ea32e86

    _Ptr->_Prev = nullptr;

• <array>, <tuple>: We don't need return after calling _Left.swap(_Right);.
• <chrono>, <complex>, <format>: Use direct-init instead of Type var = Type{init}; which repeats the Type unnecessarily.
• Several headers: Detach comments.
  • As usual, this isn't attempting to be comprehensive. I just noticed a few cases of ugly wrapping (including when function parameters are unnecessarily wrapped).
• <memory_resource>:
  • Use centralized _STL_INTERNAL_STATIC_ASSERTs instead of user-visible static_asserts before every static_cast. (This is internal machinery, so we're just checking our own work here. I believe that this code may have predated the introduction of _STL_INTERNAL_STATIC_ASSERT.)
  • Also slightly update the _Intrusive_list comment for consistency with _Intrusive_stack.
  • In _Remove(), we don't need addressof() for raw pointers.
  • Rename _Foo (a meaningless name, which we conventionally avoid, especially in product code) to _Pnext (which is used in similar loops, and is actually somewhat descriptive here - except at the very beginning, it is pointing to a _Next).
• <experimental/filesystem>: Drop // TRANSITION, ABI, #if 0 code.
  • IIRC, attempting to use _Path_cvt<value_type, char32_t> resulted in compiler errors, which we had hopes of someday fixing, or which perhaps were actually fixed by the codecvt_ids overhaul. In any event, there is no reason to attempt to enable these functions now. In vNext, we will simply eradicate <experimental/filesystem>.
• Several headers: Construct temporary strings with braces.
  • Note that string_type(1, _FS_PREF) cannot be changed, as that would attempt to construct a sequence of 2 characters (guess how I found out):

    STL/stl/inc/experimental/filesystem

    Line 1086 in ea32e86

    return path(string_type(1, _FS_PREF));

[StephanTLavavej]

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.