[PATCH 3/3] Implementations for exp(float) and exp(double)

Use separate implementations instead of a macro
to ensure the constant multiplied with is of
higher precision.