added calcAmpSum backend

[TysonRayJones]

Example PR A

Sub PR 2

This PR implements the backend of the calcRealAmpSum function from PR #607, to demonstrate how accelerated functions are structured.

Note the backend functions actually implement calcAmpSum which returns both the real and imaginary components of the amplitude sum. This will be leveraged by a subsequent calcAmpSum() API function.

In general, a function in quest/src/api calls a

1. localiser.cpp routine which communicates data between distributed nodes as necessary.
   It does this by invoking one or more routines within...
2. comm_routines.cpp
   then proceeds to invoke an
3. accelerator.cpp routine which decides whether to invoke CPU or GPU code.
   This calls either a...
   • cpu_subroutines.cpp routine which may use multithreading, or a...
   • gpu_subroutines.cpp routine which will invoke either a custom kernel, a Thrust routine, or a cuStateVec routine. These three types of functions are respectively defined in:
     • gpu_kernels.cuh
     • gpu_thrust.cuh
     • gpu_cuquantum.cuh

In this PR, the GPU backend for the function calcAmpSum was implemented using Thrust since it trivially mapped to an existing Thrust routine.

Note that calcRealAmpSum can be called upon both a statevector and density matrix Qureg. Since the simulation logic is agnostic to either, both types of Qureg were passed to backend functions containing the _statevector_ infix. If the statevector and density matrix logic had differed, we would additionally define backend functions

localiser_densmatr_calcAmpSum();
accel_densmatr_calcAmpSum();
cpu_densmatr_calcAmpSum();
gpu_densmatr_calcAmpSum();
thrust_densmatr_calcAmpSum();

and invoke the appropriate localiser function at the top-level:

qreal calcRealAmpSum(Qureg qureg) {
    validate_quregFields(qureg, __func__);
    validate_quregHasEvenNumQubits(qureg, __func__);

    qcomp value = (qureg.isDensityMatrix)?
        localiser_densmatr_calcAmpSum(qureg);
        localiser_statevec_calcAmpSum(qureg);

    return std::real(value);
}

With the frontend and backends defined, the next step is to implement the unit tests like demonstrated in #609.