Add bounds support in SPSA about qiskit-algorithms

[Cryoris]

SPSA is an approximation of gradient descent and generally does not support setting bounds on the objective function. Interpreting bounds is not straightforward, for example bounds could be

• periodic: re-enter the valid area on the other side
• reflecting: re-enter on the same side
• clipped: significantly increase the value of the objective function outside the bounds

-- and each choice impacts the behavior of the optimizer. I suggest you implement a small wrapper of SPSA that implements the bounds of your choice, e.g.

# bounds are set as [(lower_0, upper_0), (lower_1, upper_1), ...]
def spsa_with_periodic_bounds(fun, x0, bounds):
    def fun_with_bounds(x):
        # wrap x into [a, b) by transforming it to x % b + a
        x = [x_i % bounds[i][1] + bounds[i][0]]
        return fun(x)
    result = SPSA().minimize(fun_with_bounds, x0=x0)
    return result