Multiple Evaluation Modes with Operator Collections

[averyparr]

Summary

1. OperatorCollections provide a modular way for Model classes to use and easily switch between different evaluation modes.

   a. Easier to implement new evaluation modes without cluttering every Model class

   b. Allows Models to work in four steps, each of which is modular and separated from the main Model implementation.

2. Rotating Wave Approximation is now handled by a function mapping Model classes to other Model classes

   a. Clearer interface on the user side with easy signal handling for translating pre-RWA to post-RWA.

   b. Core functionality is implmented in helper functions, which makes future RWA implementations easier

   c. Most of the perform_rotating_wave_approximation function is simply casework to determine which type of model it is
   dealing with, along with a smal bit of special handling for each Model class. This makes adding new Model classes
   relatively straightforward.

3. Additional features added over the past few weeks

   a. Non-vectorized LindbladModel evaluation and RotatingFrame support for LindbladModels.

   b. Basic sparse matrix operator storage and calculation (csr_matrix)

   c. Rotating frames & basis transformations for vectorized evaluation are implemented in a manner that's simpler to the user
   (e.g. passing only $F$, and the Frame object figuring out how to deal with vectorized operators).

   d. Diagonal frame operators no longer waste time changing bases.

   e. Optimizations to speed in a few existing places, as well as several rounds of optimization within OperatorCollections.

   f. evaluate_hamiltonian functionality for LindbladModel

Details and comments

[CLAassistant]

All committers have signed the CLA.

[DanPuzzuoli]

This is good to merge.

For records: while there are no comments on this PR directly, it has been heavily reviewed in a pre-existing PR (which is now closed), as well as communication outside of github.