[ROM] Unify ROM solver

[rubenzorrilla]

📝 Description
Similar to what we did in #9439 with the RomAnalysisStage, in this PR I'm unifying the ROM solvers in a new RomSolver. The dynamic inheritance mechanism is equivalent to that of #9439. In order to keep backwards compatibility I kept the old solvers (adding a deprecation warning) and created a new_python_solvers_wrapper_rom.py module. This module filters the base solver from which the ROM solver has to inherit.

In this regard I'd like to note that, if we had a unified name for all the Python solvers wrapper (e.g. python_solvers_wrapper instead of python_solvers_wrapper_fluid), the new ROM wrapper would be simplified and easier to maintain. In my opinion, it makes not so much sense to have different names for the Python solvers wrappers among applications if we follow the rule that one application should have only one "standard" wrapper (I said standard because of the adjoints). In any case this is to be discussed in an independent issue.

This requires #9467 to be merged first.

[sunethwarna]

I have a small question regarding the GetDofsList method. If this is becoming the standard, can we use this method to correctly define the dofs rather than using model part elements to get the dofs list. So then we can simplify AddDofs method to use it?

[rubenzorrilla]

memo Description Similar to what we did in #9439 with the RomAnalysisStage, in this PR I'm unifying the ROM solvers in a new RomSolver. The dynamic inheritance mechanism is equivalent to that of #9439. In order to keep backwards compatibility I kept the old solvers (adding a deprecation warning) and created a new_python_solvers_wrapper_rom.py module. This module filters the base solver from which the ROM solver has to inherit.

In this regard I'd like to note that, if we had a unified name for all the Python solvers wrapper (e.g. python_solvers_wrapper instead of python_solvers_wrapper_fluid), the new ROM wrapper would be simplified and easier to maintain. In my opinion, it makes not so much sense to have different names for the Python solvers wrappers among applications if we follow the rule that one application should have only one "standard" wrapper (I said standard because of the adjoints). In any case this is to be discussed in an independent issue.

This requires #9467 to be merged first.

See my last comment regarding the GetDofsList (PR #9467).

[philbucher]

I recommend coming up with a better name rather than prepending with new_
remember how long we were stuck with the new_linear_solver_factory 😅

[rubenzorrilla]

😄 Considering that this is a small research application I don't think we'll need to change keep backwards compatibility for a very long period of time. Indeed, the plan is to remove the old stuff ASAP. I'd keep the new_ to constantly remind us to do it 😄.

[Rbravo555]

LGTM now