Simplify and document generation of (per-site) descriptors about acepotentials.jl

[cortner]

evaluate_basis(basis, args...) doesn't work for me. The fact it's a basis we are evaluating is already encoded in the argument. From the discussion on Slack I suggest:

• evaluate(basis, atomic_env) : the current evaluate
• evaluate(basis, structure) : the current energy
  And then, energy would just become an alias.

Next - do we need an alias for evaluate that sounds more like "evaluate the descriptor"? The site_descriptor could just be achieved via

evaluate(basis, at, 1)

Instead of additional syntax for evaluating all the per-site descriptors we could allow

evaluate(basis, at, [2,3,5,10]) # for a few site descriptors
evaluate(basis, at, :)  # for all site descriptors.

But maybe this is getting too cryptic now and the site_descriptor and site_descriptors syntax will be clearer. (Or could again be aliases...)

[wcwitt]

I added a very basic tutorial addressing this a while ago: https://acesuit.github.io/ACE1pack.jl/dev/literate_tutorials/descriptor/.

So I vote we close for now and revisit when necessary?

[cortner]

what about the terminology, that's not really settled yet.

[cortner]

@wcwitt -- A few people at UBC that are starting to use ACE as features for classification and data analysis tasks. So I'd like to come back to this, make this functionatlity convenient to access and then close this.

Maybe site_descriptors and output a matrix of descriptors is the best choice? And site_descriptor(..., i) which is now site_energy so just an alias for better readability.

[wcwitt]

The first version is done - we can revisit if necessary