Update train.py

[mhdmyz]

I highly suggest for simplifying and easier debugging and better readability and understandability of the codes, the environment creation gets it's own function and gets separated from the train and test functions. One of the prominent advantages of FinRL-Meta is it's modularity and having different independent layers, which the proposed change helps going further in this direction.

[zhumingpassional]

in the train function, how to init the dp = DataProcessor?

[mhdmyz]

The DataProcessor is initialized in the create_train_env function which results env_train which will be given as an input to the train function.