Multiple parallel optimizations

[ChristianZimpelmann]

Run several optimizations in parallel (args, criterium etc. are then entered as a list)

First: all relevant arguments need to be entered as a list.

[ChristianZimpelmann]

The code could be a bit shorter if also single optimizations are broadcasted (to lists of length 1).

Then this part is not needed:

if n_opts == 1:
        # Run just a single optimization
        result = _single_minimize(
            criterion=criterion,
            params=params,
            algorithm=algorithm,
            criterion_args=criterion_args,
            criterion_kwargs=criterion_kwargs,
            constraints=constraints,
            general_options=general_options,
            algo_options=algo_options,
            dashboard=dashboard,
            db_options=db_options,
        )

However, I think it is clearer to make this distinction and don't initialize multiprocessing if it is unnecessary.

[janosg]

Yes, we should keep single optimization as special case!

[janosg]

The changes made the code much more readable. Just two questions left, then we can merge. Thanks!

[ChristianZimpelmann]

The changes made the code much more readable. Just two questions left, then we can merge. Thanks!

I took care of your remarks and simplified the code once more (the broadcasting function isn't needed anymore)

Also n_opts is now just len(arguments). So I suggest to not return it separately.

[janosg]

Very nice. Thanks a lot for this PR!