Lazy load backends on the cloud channel

[kt474]

What is the expected feature or enhancement? https://github.com/Qiskit/qiskit-ibm-runtime/pull/791 added lazy loading for the ibm-quantum channel. For consistency, It would be nice to have this on the cloud channel too even though the performance improvements don't appear necessary

Acceptance criteria

[jyu00]

We should still do this as we are adding more 100+ qubit backends to the paygo channel, as well as moving more workloads there.

[daka1510]

@kt474, @jyu00 any updates on the priority of this issue? I've just hit this problem that blocked me running a job on device_x because the initialization of device_y failed

https://github.com/Qiskit/qiskit-ibm-runtime/blob/9fc853b0ba301adc2308ffa1b685689cdd263e89/qiskit_ibm_runtime/qiskit_runtime_service.py#L152-L153

I agree it's not a problem (except for performance reasons) if all devices are functional, but as soon as one device is broken for some reason it's odd that it also blocks job execution for all other devices.

[kt474]

@daka1510 I can work on this next sprint, do you happen to have an error log?

[daka1510]

@kt474 no. I got an exception stacktrace that lead back to this line in the described scenario. It happened because https://us-east.quantum-computing.cloud.ibm.com/openapi/#/Backends/get_backend_configuration returned an error for one of the devices that got returned from the list endpoint.