add bonds to AbstractSystem

[ejmeitz]

Added bonds function to AbstractSystem and an example of a BondedSystem. The example is basically just a FastSystem. I have not added tests to this yet.

This is based on discussion from #87.

[rkurchin]

Left some comments, focusing on the interface itself for now. Once we have consensus on that, we should also make sure to add some tests for the BondedSystem type.

[rkurchin]

Can you add a default implementation that just returns an empty tuple? e.g. instead of this line, something like:

bonds(sys::AbstractSystem) = [()]

(I'm not sure if there's a type stability issue here with that return type not having the type parameters of the Tuple, though)

[rkurchin]

I don't think the example at the end of this line is actually in line with what you want, since the third element should be a BondOrder enum...

[rkurchin]

Strictly, Integer should probably be typeof(firstindex(sys)) (in practice, I think all implementations do use integer indices right now, but this is more general)

[rkurchin]

should probably do this as

@enum BondOrder begin
    single = 1
    double = 2
    triple = 3
end

because it seems that if you don't do that, it actually assigns them numbers in a zero-indexed fashion (super weird), so the following can happen, which would be very confusing:

julia> BondOrder(2)
triple::BondOrder = 2

Separately, there might be a longer-term discussion about whether we should break out bonding types to be part of the interface since I could imagine people wanting more flexibility here, but my opinion is we start with this and file an issue for future conversations about this.

[jgreener64]

My suggestion would be to not have the bond order as a required part of the interface. Without this it becomes a more general interface for storing atom pairs.

Instead we could have properties of bonds like we have properties of atoms, and some way to ask for the properties when calling bonds. This would also allow the storing of other properties if desired.