Allow for Python flux connections, even if they are slow

[philippkraft]

The connections in cmf are written in C++ to use the full potential of speed. However, when new connections are developed, the write, compile, link, install, debug cycle is clumsy. Using Python flux connections would speed up development and can be translated into C++ when the math is working. This would also be a way to include Python developers more easily in the extension of cmf.

Proposed usage:

class CustomConnection(cmf.flux_connection):
    def __init__(self, left, right):
        super().__init__(left, right, type(self).__name__)
        self.left = left
        self.right = right

    def calc_q(self, t):
        return self.left.volume - self.right.volume

p = cmf.project()
t = cmf.Time()
s1 = p.NewStorage('S1')
s1.volume = 1.0
s2 = p.NewStorage('S2')
con = CustomConnection(s1, s2)
s1.dxdt(t)

[philippkraft]

Due to the protected status of calc_q, this is not possible for flux_connection. However with a new class BaseConnection which exposes calc_q it works.

namespace cmf { 
    namespace water {
        class BaseConnection: public flux_connection{
        public:
            BaseConnection(cmf::water::WaterStorage::ptr left, 
                           cmf::water::flux_node::ptr right, 
                           std::string type)
            : flux_connection(left, right, type) 
            {}
            
            real calc_q(cmf::math::Time t) override    
            {
                throw std::runtime_error("BaseConnection.calc_q needs to be overriden by child class");
            }
        };
    }
}