Isolated at en

[FNTwin]

Various stuffs:

• Solve Option for changing cache location #15
• Solve NablaDFT dependency #13
• Solve Dummy class for testing purposes #12

Other things:

• Implement the correct unit for every dataset
• Implement the correct level of theory
• Implement the correct force sign
• Implement the mkdocs serve with a simple doc
• Clean the env.yml and remove unwanted deps
• Implement a utility method to save extend xyz files
• Implement all the isolated atom energies and wrapped them into a nice and clean dict
• Implement factory method to automatically get the correct energy from the class
• Changes the init of the base class for the cache selection
• Better inheritance in the child classes
• Implementation of lazy loading of the modules
• No more NablaDFT HamiltonianDatabase initialization everytime you call the dataset module

To do :

• solve Unique Atom Types and Charges #5 in a bit
• correct way to apply isolated_atom_energies (right now we are not considerating the charge and the format for the dict is Tuple(str , int) as (Name of the element , charge) )
• solve the merge issue

[prtos]

why do we set all atomic energies to zeros here? don't we have them calculated?

[FNTwin]

The current atomic energies workflow implemented in main is wrong as we are not getting the isolated atom energies considering the formal charge of the atom.
Right now the energies are indexed by the chemical number but this is not taking account that elements with different charges have different energies, so we should be indexing by a Tuple(Chemical species, Formal Charge)

This is a workaround to continue using the package while I take care about it

[FNTwin]

Implemented a way to get the correct e0 considering the formal charges. See 928716b

[FNTwin]

Implemented:

• Docstrings
• Docs + tutorial + some API on mkdocs
• SOAP vectors calculation for dataset chemical space