create FCN model class

• implement a simple FCN
• implement a FCN class that can use parameters from outside
• enable parameter parsing to model class
• FCN can use tanh, linear, sigmoid, relu, prelu, selu activation functions
• FCN can use adam and sgd optimizer
• FCN can use dropout and alpha_dropout (for selu activation)
• FCN can either set up with number of layers n_layers and number of neurons n_hidden or by using the parameter layer_configuration (a list of int indicating the number of neurons for the specific layer) while usage of layer_configuration is always prefered if provided.
• FCN has parameter activation (used in all hidden layers) and output_activation (used in the output layer)
• FCN can use l1 and l2 regularization with the regularization argument, kwargs of the regularizer can be parsed too using l1= ... , l2= ... .