test metadata experiment is good grouping

Created by: wilsonmr

I wanted a test so that we know that grouping by experiment is ok*, so I wanted to get the systype overlap of all commondata

*when it comes to constructing covmats/generating pseudodata etc.

This meant I wanted to collect commondata across the groups, and I wanted this to be independent of theoryid so I added a couple fo collection rules. This necessitated expanding upon #992 to also make each data_input of group_dataset_inputs_by_metadata a NSList, whilst I was at it I made produce_group_dataset_inputs_by_experiment just call that code with fixed metadata_group

I made print_systype_overlap a little neater (in my opinion) since I was quite inexperienced when I first wrote that.

This change also meant updating test_fitdata since the inputs for the function were different and we can't use the API to test that function.