Do we want to keep the channel mask in the acq info?
If we go with the "everything has the same >5 dimensions", it would be one of the parameters requiring an exception.
If we keep it, should we maybe convert it to a boolean tensor? currently, the 1024 channels are saved as 64 uint16 tensors... 1024 bool values would be more usefull