ready to merge zheng_dev to dev

[JohnWangDataAnalyst]

1. state structure by adding one dimensional for population
2. fitting var_inv instead of var
3. model fitting add warmupwindow between epochs for training
4. mutilmodal fitting example for concurrent eeg and bold
   using two model: one is integration with nmm and eeg and bold in the same model
   the other is a model which is built on combination of nmm eeg and bold
5. authorship
6. move common codes or methods from each model into abstract class
   only difference with current dev is the class recording. current version the simulated output timeseries are saved in the statetraining