Command iterator confusing linear indexing behaviour

[ss2165]

Currently the Circuit trait is implemented on anything that implements HugrView.

That means you can ask for a CommandIterator over any Hugr. The logic is that any input linear wires go in to the indexable array, and CircuitUnit::Linear is used to refer to operations that update these linear operations.

Command of op on CircuitUnit::Linear(i) really means op takes the Wire corresponding to i and returns a Wire, and from now on i will refer to that new wire. This is useful for qubits because of how we are used to building quantum circuits, so I wonder if it is actually helpful to do it in general and not just for qubit types (and specific input qubits at that, like how the CircuitBuilder does it rather than all of them).

Further, if the op doesn't return a linear wire for every linear wire it took in (e.. MakeTuple) the current logic is that the previous indices no longer appear in commands and a new index is assigned for the new linear wire. I ran in to an issue where the tuple was then unpacked (and so in theory you could go back to using the old indices) but all the indices had now been shifted.

This behaviour is unintuitive, I suggest instead that the command iterator either terminates or errors when it cannot continue tracking every linear wire it has been charged with.