Bin-Shuffling: reimplemented the continuos time version

[pbouss]

The continuos time version of bin-shuffling preserves now the number of spikes. That is in contrast to how it was done before, where the spike train was first binned to a bool array.

[dizcza]

You're giving quantites a hard time by throwing all the units you have in a mincer to chew them later on.

A bitter example of your attitude

        surrogate_spiketrain = \
            (surrogate_spiketrain * pq.s).rescale(spiketrain.units)

cries for help.

Instead, quantities require love: do gentle stripping at the entrance, and don't forget to dress them up before you're done.

[pbouss]

You're giving quantites a hard time by throwing all the units you have in a mincer to chew them later on.

A bitter example of your attitude

        surrogate_spiketrain = \
            (surrogate_spiketrain * pq.s).rescale(spiketrain.units)

cries for help.

Instead, quantities require love: do gentle stripping at the entrance, and don't forget to dress them up before you're done.

Yes, sure. I would not consider myself as a friend of quantities. I like to work only on magnitudes, thus fast numpy arrays and only in the very end got back to quantities.
But it actually appears much more natural as you did it now. Thx.

[coveralls]

Coverage decreased (-0.7%) to 89.606% when pulling 1262dab on INM-6:feature/continuous_bin_shuffling into 4a72095 on NeuralEnsemble:master.