Standardize features across days

[lukovkin]

Right now we standardize for each day separately.

Approaches:

• Brute force: try to fit data from all files into memory and perform standardization (could be unscalable)
• Calculate means and stds for each file separately, then calculate 'general' mean and std, then apply 'general' mean and std to the each file
• Perform it on Spark/Hadoop?

[lukovkin]

Done.