Update simulation scripts

[Tony911029]

1. Remove unused columns during data generation
2. Use a random seed for each patient
3. Truncate data such as bgl values to save some spaces
4. Remove some extra arguments for run_glucose_simulation
5. After each patient's data is generated, save it right away instead of caching in a dict to save some mem space
6. Clean up side products from the simulation such as performance metrics (not relevant to us)
7. Add id to simulated data for benchmark repo

Some new patient stuff:

1. Add new params for new patient (adults for now and max is used rather than mean or median)
2. Add a notebook to restore the values of state variables for RL

TODO: Merge the notebook with the one that generates the param to make the repo cleaner

[RobotPsychologist]

Looks good!