Storing quantities of interest as dataclasses
Description
I propose we turn the quantities of interest (dictionary) into a (json-serializable) dataclass that is unique for each particular analysis. I also propose we do not store the fit-results as part of this object.
Motivation
Currently the contents of the quantities of interest is hard to predict, somewhere in each analysis class random methods write to this object. Turning this into a dataclass would allow us to document what all the fields are and ensure it is possible to rely on the contents of these objects when building other functionality that relies on analysis.
For reference, the issues workflow is described in the contribution guidelines.