Inconsistent use of the new Hamiltonians interface

[mtsch]

When preparing PRs #331, I ran into a major issues with the new Hamiltonian interface. This is the situation:

• A majority of our tests (including doctests) still use the old interface. If a Hamiltonian is updated to the new interface, a lot of work needs to be done to fix them. The docstrings showing the old interface (see for example most doctests in Interfaces/hamiltonians.jl) must be confusing for new users as well.
• Some code still uses the old interface. I tried to fix this where I could, but I think there are still cases I missed.
• Updating a Hamiltonian to the new interface breaks user code that still relies on the old interface making what should be a non-breaking change breaking.

To fix this we should either create fallbacks for the old interface that creates an operator_column in the background or eliminate all traces of the old interface from our code.

The first option is non-breaking and easy to do, but will result in some circular definitions. The second option is a lot of work (since we'd need to update all of our Hamiltonians) and breaking, but will result in less convoluted code and may be less confusing to new users.

@joachimbrand @jamie-tay thoughts?

[joachimbrand]

I would prefer to move everything to the new interface as soon as possible. We will need to identify where this is breaking. Ideally I like changes to be incremental but if we need a bigger change then so be it.

[joachimbrand]

I'd also be happy with the first option as an interim solution, if it is easy to do, and then plan the full transition for the next PR. However, this may turn out to be more difficult than removing all traces of the old interface.

[joachimbrand]

Is this still an issue or do we still have Hamiltonians in Rimu.jl that are not compliant with the new interface (and would not work with ModifiedHamiltonian)?