sample implementation of a (uncorrelated) combined fit

[PiaLJP]

A combined fit is performed when dictionaries (instead of lists) are handed over to the fits.least_squares function.

[fjosw]

Hi Pia, it looks like your PR breaks a few tests. Please have a look at CONTRIBUTING.md where I tried to explain how to run the tests on a local machine.

[s-kuberski]

Thanks for the implementation. Before I inspect everything in detail, I'd like to ask a general question: Is it really necessary to do the combined fit in a new routine that has many common aspects with _standard_fit? I have the feeling that this will increase the amount of work that will be needed in the future to implement new routines and to fix and avoid bugs.

My first attempt would have been to write a wrapper around _standard_fit or to adapt the first lines of _standard_fit such that proper combinations x, y, func are handed to the routine. For x and y this is trivial, for func one could just settle on a list of functions that is filled with the same function over and over again if no combined fit is done. Did you look at this and decided to not to do this for some reason (performance?)?

For the future of the fit routines in pyerrors, I'd rather merge all the fit routines (i.e., also _prior_fit would be considered here) instead of adding new routines.

[fjosw]

I agree that this is not a good solution and results in a lot of unnecessary code duplication and maintainability problems. I see this PR as a step towards a new version of the fit module as it already defines the interface for combined fits and provides test cases which will also have to be fulfilled by the final version which could be very helpful in the next development phase.

In this spirit it would be very interesting to hear your opinion on the interface @PiaLJP designed and the overall functionality, not so much the actual implementation. As soon as we agree on the interface we could then start to rewrite and refactor the code.

[s-kuberski]

Thank you very much for your effort. Let's go with the dictionary solution by now. In contrast to the other solutions that I had in mind, it would remove the need to have x, y and func in the same order, if the implementation is adapted accordingly.

[s-kuberski]

Hi. Thanks for adapting the code. I have made one minor comment that we could resolve before merging.