New entropy functions and dump_colored_trj.

[matteobecchi]

Requested Reviewers: @andrewtarzia @SimoneMartino98

Two improvements:
First, I added the method Trj.dump_color_trj() which takes an array of labels and saves a .xyz file with the atoms correctly labelled. There is an analogous function for OnionSmoothInsight, but in this way the user can do it with any array of labels.

Second, I created two new functions to compute Shannon entropy and information gain, respectively. The idea is that these functions are better than the many functions implemented before, so the user should stick to those.

They are better in the sense that:

• They work for bot uni- and multivariate datasets (no need to distinguish between "shannon" and "shannon_multi" and stuff like that)
• They work for both discrete and continuous observables, and the user always has to set the nature of the dataset he is computing the entropy of. In this way, users should be "encouraged" to compute entropy in the right way.

I added to the old functions a warning saying they are deprecated, then I left them there for backward compatibility.
I updated the examples and recipes accordingly.

[matteobecchi]

I started to make the names of parameters in functions and methods consistent (see issue #125 ).

I did the easy part, what remains to be done requires to modify the lens functions, it will be a future PR.

[andrewtarzia]

Great work @matteobecchi

[SimoneMartino98]

All good, Approved