Add a first option for running imaging as an optimization problem. OptimizationHandler derives from SynthesisImager, and implements both the chi2 evaluation (in visibility space), as well as its gradient (through the major cycle).
Python optimization could now induced by, e.g.:
optim = SynthesisOptimizationHandler(paramList)
x = optim.x0.copy()
for i in range(100):
x += 1e-5 * optim.gradient(x) #gradient descent
print(optim(x))
or more sophisticated through scipys optimization framework. Note that the chi2 function is also the Bayesian likelihood, so sampling with the help of python MCMC should also be possible, as well Plug-and=Play AI denoising. For RML and sparse imaging, one needs to add the regularization function (and its gradient) in python, and add functionals, e.g. by lambda framework.
Most desirable would be an integration into a framework such as regpy.
This is not the final version and likely to change.
