FloatingPointLiteral: Constructor for special consts. (#6220)
This replaces static helpers for creating special consts with
constructors. This is in preparation for a FP literal implementation
using MPFR. In the next step, I'll introduce a FloatingPointLiteral base
class, from which the specialization FloatingPointSymFPULiteral is
derived. The MPFR implementation will also be derived from the base class.
This is in order to make unit tests that compare between the two
possible. Further, in the worst case, MPFR will have to use SymFPU for
unsupported cases (to be determined).