SFT's hidden _data should never be used.

[EmmaRenauld]

As discovered with François, the sft.data_per_point[key]._data is not always updated when shuffling or modifying the SFT.

This is probably true for all other usages of hidden _data.

I cleaned one line of code (see #890) , but code should be verified everyhwere.

@frheault

[arnaudbore]

is it only dpp and dps ? or every single one like sft.streamlines._data ?

[frheault]

It is true of that one too, luckily Scilpy scripts are usually doing the right thing, but this is very dangerous when modifying a script, developing a new feature or hacking away in IPython.

The array sequence use something similar to https://numpy.org/doc/stable/reference/generated/numpy.ndarray.view.html

This means that if you slice an array sequence you are playing with a view and the underlying data is untouched