Add batch_size to nplet measures and heuristic functions

[Laouen]

the following functions comput in a single batch the given input:

nplet_measures
nplet_measures_hot_encoded
greedy (the repeat parameter is the batch)
simulated_annealing (the repeat parameter is the batch)
simulated_annealing_multi_order (the repeat parameter is the batch)

The problem is that in some cases the single batch could be to big (for example if repeat == 2000) or if doing an analysis taking all nplets with some attributes. Then, we need to implement a dataset and dataloader to process in smaller batches where the user controls the batch_size as parameter.

[Laouen]

This feature is becoming very relevant for easy use of the tool in not powerful devices