Handle "inf", "-inf" and "nan" for double and float types

[darcamo]

Currently when parsing values as double or float, cxxopts uses the fallback parser using stringstream. However, using input streams does not support a representation for infinity and NaN (see this stack overflow question) .

It seems that all that is needed to support passing "infinity" and similar values that can be properly parsed is to add two specialization of parse_value, one for double, that uses std::stod, and one for float, that uses std::stof.

Another specialization for long double could also be added that uses std::stold.

Note that std::stod, std::stof and std::stold were introduced in C++11.