This function uses the functions mentioned below to write a stand-alone C
program containing all the weights and biases (theta's) of net. Also
included is a routine eval_c_ffnet (see print_c_code) to evaluate that
network.
The network written has fixed dimensions.
NOTE
The network must have type FFNET (see NET-flags) and contain 3 layers.
Furthermore, the activation function used must be ActInprod and the
transfer function either TransLinear or TransSigmoid for output
units (but equal for all output units) and TransSigmoid for hidden units.