Update net.py, add num_layer to network class

When creating a neural network, we can specify the network width, but we can't set the number of layers, and I want to add the hyperparameter `num_layer`. To be compatible with the previous code, the `num_layer ` is set to 3 by default.

创建神经网络时，我们可以指定网络宽度，但无法指定网络层数，我想要为网络增加 num_layer 这个超参数。为了兼容以前的代码，num_layer 这个超参数被默认设置为3.

Do you have any other suggestions to change it?

请问大家有其他修改建议吗？

For example, use hyperparameters like `net_dim=(512, 256, 128)` to set up a three-layer neural network with 512, 256, 128 features in each layer, like ray-RLlib does. But I haven't meet the need to "set different network widths for each layer" yet.

例如使用 (512, 256, 128) 这样的超参数去设置一个三层的神经网络，每一层的网络特征数量分别是 512， 256， 128，就像 ray-RLlib做的那样。但我自己还没有遇到过“需要为每一层设置不同网络宽度”的情况。

Start with net.py and the network class to trace how the current width and layers are constructed. Add the num_layer hyperparameter with a default of 3, then verify that existing network creation remains compatible and that a different layer count is honored.

Escrito por el modelo de indexación a partir del texto del issue.