Builtin DFP types for C++ using `__attribute__((mode(SD/DD/TD)))`

[tahonermann]

The DFP support defined for C++ via ISO/IEC TR 24733:2011 does not include support for builtin DFP types or the _Decimal32, _Decimal64, and _Decimal128 keywords; support for DFP types is instead relegated to the standard library. Gcc implements the specified std::decimal::decimalNN types as light wrappers around builtin types that are equivalent to the _DecimalNN types but that lack keyword names. These types are instead named using type attributes as follows:

• float __attribute__((mode(SD))) # _Decimal32
• float __attribute__((mode(DD))) # _Decimal64
• float __attribute__((mode(TD))) # _Decimal128

This issue tracks doing the following:

• Adding support for the above type specifier attribute syntax and mapping it to the builtin types added via issue ​AST representation for builtin types #4.
• Disabling the _DecimalNN keywords when compiling as C++ (in clang/include/clang/Basic/TokenKinds.def, the _DecimalNN keywords are defined with KEYALL; this should be KEYNOCXX).

[tahonermann]

Completed via #23 and commit ef6273c1e72f4b57810a4fa661165fb7c9ecc393.