Rewrite models based on Square struct, add LineMPO struct

[LinjianMa]

Change the hamiltonians in the PEPS optimization experiments such that each local hamiltonian is a column/row of the overall H.

[mtfishman]

This is useful functionality!

In general what we could do is make this more generic, where you could ask for a general (hypercubic) slice of a HyperCubic lattice type and then ask for the bonds of that slice. But totally fine to have more specialized functions like this for now.

[mtfishman]

Looks good, glad you were able to fix the bond iteration for the HyperCubic lattice type.

I think ideally we wouldn't have two different types LocalMPO and LineMPO, in principle we can merge them into one type where it stores the list or range of sites where that Hamiltonian term is acting. But the current design is fine for now.

We should make issues reminding us to:

1. Merge LocalMPO and LineMPO into one type.
2. Generalize the version of bonds that returns an iterator over the bonds of a hypercubic sublattice of a hypercubic lattice (and consider writing a general SubHyperCubic type which is a slice of a HyperCubic type, or allow adding slicing information into the HyperCubic type itself).