Num_π

π (18 digits)

Unlike the original BCDFloat from Kai G. Schwebke I use a fixed maximum precision defined at compile time - so there is no need for truly arbitrary irrational constants.