[Draft] operator API revision to align with specs

[bettinaheim]

I still meant to move the operator files into a separate folder. I'll do that last once I've checked everything else looks good.
I'll also take another look to exhaustively list what really needs to be breaking changes, and what can be further mitigated.

Most of the changes are just renaming + updates to the new API to avoid using what will be deprecated. The meaningful changes are:

• operators all properly track identities; that means that spin_op::x(10) is no longer considered to be acting on 11 qubits... This is the heaviest breaking change we can't get rid of - it needs to be broken. It is just plain wrong that an operator on a single 2-level degree of freedom does not produce a 2x2 matrix
• Still to be contemplated: The existing default constructor for spin_op also has the oddity of adding a 1. This too imo is just plain wrong; we see plenty of places where a default was constructed, terms were added, and then the 1 is subtracted again. It also assumed it was acting on degree 0. This default constructor needs to go. The intended constructor for sum_ops is indeed an empty sum. however, because nothing times something is still nothing, that might be surprising and would be breaking existing code pretty heavily. Correspondingly, I made the default protected and added an explicit ::empty (= mathematical 0) and ::identity (= mathematical 1) for construction. Not having a default constructor though would also break code, and we could properly deprecate it...
• What is done in the simulator files
• Unification of complex_matrix and matrix_2, now called complex_matrix
• Align degree vector order with the order that simulators expect
• Changing the key that the cudaq runtime uses to retrieve expectation values for spin ops
• two or three minor changes in the signatures of internal functions (spin_op_term vs spin_op, adding missing const)
• adds proper iterators for product op and sum op, and removes get_terms
• a bunch of class renaming as requested by the spec

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[bmhowe23]

This is a different order than the order called out in the comment on line 27. Does that matter? Should the comment or enum be updated?

// I = 0, Z = 1, X = 2, Y = 3

[bettinaheim]

The order does not matter, and neither ever does the private op code - it is private for a reason. :)
That said, I can align it with this one so that the data format and the internal representation match.

[bettinaheim]

/ok to test

Command Bot: Processing...