fix filenames and task ID printouts when loading (human) demonstrations from a larger dataset than num_train_Tasks

[tomsilver]

for some reason, the current filenames for results are not even including the number of train tasks when we're doing:

python src/main.py --env stick_button --approach nsrt_learning --seed 0 --option_learner direct_bc --segmenter contacts --demonstrator human --load_data --min_perc_data_for_nsrt 1 --num_train_tasks 1000 --timeout 300 --neural_gaus_regressor_max_itr 50000

and with different values of --num_train_tasks