✨ `BASEURL` support through QDMI authentication

[marcelwa]

Problem Statement

#1355 introduces support for QDMI authentication via FoMaC. However, it doesn't currently offer a way to provide the BASEURL parameter because this is a DEVICE_SESSION_PARAMETER, not a SESSION_PARAMETER.

Proposed Solution

I don't know a good solution for this problem yet, but it will be closely related to solutions for #1361 and #1362.

[burgholzer]

This relates to https://github.com/Munich-Quantum-Software-Stack/QDMI/issues/161, although that issue is a bit broader in scope.

For the matter here, there currently is no way to provide a BASEURL for a device from the client side.
We could simply add such a parameter to the QDMI_SESSION_PARAMETER enum as a workaround until we figure out a proper way for clients to forward device session parameters through a driver.

However, this does more broadly relate to Munich-Quantum-Software-Stack/QDMI#215 as well and the question of whether the client interface is actually necessary. This has come up multiple times so far, and the more I think about it, the more often I think it would be beneficial to simply drop and build up on the QDMI Device interface.
This would still mean that a driver exists that loads the name-shifted shared or static libraries, but that it wouldn't have to go through a C API, again, before being wrapped in another C++ abstraction.
Dropping the interface would also allow us to directly set parameters for the individual devices, because we would be the ones in control of how the sessions are created with the devices.
At the same time, we keep thinking about this "stack-as-a-device" concept. And all the discussions so far have led more and more towards the insight, that a separate interface will probably be required for exposing stacks. Such an interface would be highly inspired by QDMI, particularly the client interface supposedly, but would be separate.

Concrete proposal:

• Drop QDMI client interface
• Merge Driver and FoMaC as C++ abstraction over QDMI
• Expose the resulting C++ API to Python (almost identical to now)
• Work on a separate Stack-Interface (QSMI) inspired by the client interface (once worked out, it would be up to the individual drivers to implement logic for dynamically loading libraries implementing the new interface)

@ystade your input would also be valuable here.

[marcelwa]

As discussed yesterday, I've extended the scope of #1355 by passing DEVICE_SESSION_PARAMETERs to Driver::addDynamicDeviceLibrary and exposing that function to Python also. This provides a functional way, albeit not the prettiest. In the long run, I'd prefer a more idiomatic solution.