Adjust DynamicsBackend level 2 result formatting to match BackendSamplerV2's expectations

[wshanks]

Qiskit Experiments 0.8.0 uses qiskit_ibm_runtime.SamplerV2 to execute circuits by default and that passes the execution to qiskit.primitives.BackendSamplerV2 for DynamicsBackend.

When BackendSamplerV2 converts the results from DynamicsBackend into the PubResult format, it uses the result memory in an if statement to check if it is empty. For a list that works but for a numpy array an exception is raised about needing to use all() to check the truth value of an array. Here the memory data is converted from numpy array into a list using tolist() before being passed into the result object. (Note: the function generating this numpy array has a type annotation stating that it returns a list, so this could be changed at that level of the stack instead but that might require changing other things expecting a numpy array).

After checking that the result memory is not empty, BackendSamplerV2 packs the state results into a BitArray and to do this it assumes that the memory results are formatted as hex strings but DynamicsBackend was returning binary strings with the format "10101". Here the strings are converted to hex. It is not clear to me if Qiskit specifies that one format is required or the other or if both should be valid (and BackendSamplerV2 really has the problem by not accepting the binary format). There could be downstream disruption if some code assumes DynamicsBackend returns the binary format. Additionally, does Dynamics support returning shots for higher levels up to 9? That would work the "binary" (really "decimal" at that point) format but forcing the casting to hex breaks it.

[DanPuzzuoli]

Thanks @wshanks , looks good to me!