Offer an async API

[Yoric]

Some of our users are operating from Jupyter, but some will be developing interactive or backend applications.

For these users, it would be nice to offer an async API, with the ability to place HTTP calls, wait for batch completion, etc. without blocking the main thread.

[Yoric]

Right now, our users are presumably mostly writing Jupyter notebooks. However, as we expand our reach and the number of libraries, we expect that some users will end up writing backends and/or GUI applications that rely on some of our libraries, hence on pasqal-cloud.

While blocking the entire application during a HTTP call is absolutely not a problem for a Jupyter notebook or a CLI application, this is not acceptable in terms of perceived performance for a backend or a GUI. So we need the ability to make our application non-blocking. Which, in Python, translates to using the async version of request to make HTTP requests and making all the functions or methods that call it async.

Note that this is a multi-repo project, as we'll also want to make changes in pulser_pasqal. This conversation is also taking place with the Pulser team.

[Yoric]

So, inspecting our code a bit more closely, it looks like there are two main blocking operations: request.request and time.sleep. I've just tested in the REPL and it looks like both of these operations release the GIL, which means that they can be executed in background threads.

Which means that if we want to have an async SDK, we might be able to get away with running it in a background thread, instead of rewriting the SDK to be async. In such a scenario, wait=True would cause issues (or at the very list, it would block the entire thread, which would severely hamper scaling), but it might be an acceptable solution for the time being.