Hyper parameter optimization for sliding window arguments enabled

• feat!: removed data preparation out of predictor class, sub_seq, steps_past, steps_future need now to be defined in each model method
  • allows for advanced hyper parameter tuning
• fix: removed tensor board activation logic bug