QUDORA integration

[Sola85]

Description

This is a first prototype of a cudaq integration with the QUDORA api by means of a new ServerHelper.

There are still a few things to do:

• Update documentation
• Implement tests using our production API (not just the mock server). This still requires an update on our side.
• Answer the questions below

Despite these open points I'm opening this draft pull request to get some early feedback.

Some open questions:

• Some kernels (e.g. ctrl_z_kernel) generate unused (and hence undesirable) __quantum__rt__read_result. Can those be eliminated automatically by adjusting the compilation pipeline?
• We would like to rely exclusively on the qir output log for passing results back to cudaq. While the existing createSampleResultFromQirOutput is sufficient to pass all current tests, the resulting shots dict seems to contain less information than the qir output log and in particular looks different than shot dicts generated with cudaqs native simulator (__global__ and named outputs are missing in the shot dicts generated using createSampleResultFromQirOutput). Is this a problem?
• Currently, we generate 1 ServerMessage per KernelExecution object, i.e. each job only contains 1 kernel for execution. However, our api also supports multiple independant kernels/qir-programs to be submitted in one "job" for a reduced communication overhead. The current api of ServerHelper does in principle allow generating only 1 ServerMessage for a std::vector<KernelExecution>, however, we saw no way of retrieving these bundled results using the current API of ServerHelper in a manner compatible with cudaq.observe. Did we miss something here, or is "multiple kernels per jobs" simply not supported currently?

In case thats easier, we are happy to discuss any of the above points (in particular how to proceed with production tests) in a call.

[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.