This is experimental.  There is no support available.  Please do not
file bug reports unless they have ready-to-commit patches.

This adds support for _Decimal64, a 16-digit base-10 floating-point
type.