Change assumed shape of time series to timepoints*channels

[SveaMeyer13]

I assumed the shape of sktime time series to be channels * timepoints (as it is for example defined in mne). But when reviewing the specifications made in PR #1232 in sktime, I figured that it is supposed to be timepoints * channels.

This required changes in

• the two series-to-series transformers
• the epoching function and
• the example notebook.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[fkiraly]

@SveaMeyer13, on this note: in the sktime branch datatypes-doc I've added an example notebook 00-datatypes which spells out the machine type specifications of series and panel data. Optimally, any in-memory container convention should comply with one of the conventions specified in there.

Now I'm aware this is the first time we've actually written this down, so any input is appreciated. Especially if you think we need a different in-memory representation of an existing scitype, or new scitypes even, to support your designs for sktime-neuro.