don't use operator++ for floating point type.

I would say that using operator++ for a floating point type is a little confusing. This is why I made this merge request.

However, if you feel that this merge request is not helpful, then close this MR and I will modify file ThinRealWrapper.hpp in !362 (merged) and add operator++ there. That file ThinRealWrapper.hpp was written to catch all incorrect usages of no-double types during compilation and to workaround boost python losing 3 digits of precision when using long double type.