optimize slicing word based signal from main signal

[hvgazula]

https://github.com/hassonlab/247-encoding/blob/encoding_with_pickles/tfsenc_utils.py#L151

Please refer to https://stackoverflow.com/questions/59238003/how-can-i-define-multiple-slices-of-a-numpy-array-based-on-pairs-of-start-end-in