[Realistic-sims, Benchmarking, Measurements] Improve performance for classical shadows method

[madagra]

Classical shadows algorithm performance can be greatly improved by doing appropriate parallelization of the loop for shadow computation and in other places within the code (see comments throughout the code).

UPDATE 06/11/23: let's use this issue for testing also the performance of the bitstring corruption routines which might be quite inefficient. Up to how many qubits can we go?

[RolandMacDoland]

Part of Ethan internship ?

[mlahariya]

@chMoussa Please review if this needs to be revisited in protocols. Otherwise we can close it.

We only need to address it if there are performance issues.

[chMoussa]

@mlahariya This will be revisited in protocols. I also have #608 open that fixes an issue with the current shadow implementation and improves a bit the runtime.