Improvements validation function and offline validation

• modify visualization functions towards latest changes in main
• create new file for more franka data specific validation
• add correct titles and description of the visualisation
• improve efficiency of visualize_recon_loss function
• train models with different window_sizes
  • for evaluating order of magnitude window_sizes = [10, 50, 100, 500, 1000]
  • keeping batch_size and epochs constant
• validate the trained models utilizing the updated validation tools
• discuss the results together, also regarding a suitable threshold for AD