Update comment

src/RiemannR.cpp

@@ -4,9 +4,9 @@
 ///         function which is a very accurate approximation of PrimePi(x).
 ///         The accuracy of this implementation depends on width of the
 ///         long double type. If the long double type has 80 bits (e.g.
-///         Linux) then RiemannR(x) is accurate up to 10^19, if the long
+///         Linux) then RiemannR(x) is accurate up to 1e15, if the long
 ///         double type has 64 bits (e.g. MSVC & macOS) then RiemannR(x)
-///         is accurate up to 10^15. This implementation could be made
+///         is accurate up to 1e12. This implementation could be made
 ///         more accurate using the non standard __float128 type, but for
 ///         primesieve's purpose speed is more important than accuracy.
 ///