Petition for flattened dictionaries everywhere

TOML allows headers, which when parsed with Base.TOML convert to nested dictionaries. For the pipeline it might it is convinient to flatten any nested dicts. This allows to collect many parameter sets into a single DataFrame.

Right now we use the nested dicts to clearly group/distinct parameters for different components from one another. We then just pass dict["Mesh"] if we only want Mesh parameters. If we flatten the dict, then we can't do that anymore. Currently we also have a separate check in place to test whether a "Mesh" section exists in a dict.

Introducing flatten dicts allows to effortlessly convert to and from DataFrames. If we add a query method which uses a regex pattern to extract the desired keys, then this could also remove the separate check for sections.