Add ChemicalDomain.from_dataset

[lilyminium]

Is your feature request related to a problem?

I manually made the SMILES strings myself for the ChemicalDomains in the published models, but this is a very automatable process if we just iterate over all possible elements and all possible bonds and check for presence in the training data.

Describe the solution you'd like

Describe alternatives you've considered

Additional context

It would be nice if we could amalgamate patterns (e.g. [#17:1]=,#[*:2] instead of manually listing every iteration) but definitely a nice-to-have rather than essential.