Avalanche Subset does not define task_set

[AntonioCarta]

The bug can be reproduced with the following script:

from torch.utils.data import random_split

from avalanche.benchmarks import SplitMNIST
from avalanche.benchmarks.utils.data_loader import TaskBalancedDataLoader

benchmark = SplitMNIST(5)
data = benchmark.train_stream[0].dataset
rem_len = len(data) - 100
buffer, _ = random_split(data, [100, rem_len])

dl1 = TaskBalancedDataLoader(data)
dl2 = TaskBalancedDataLoader(buffer)

Do we have any other mechanisms to subsample the data?