Design and document new structure for storing MIP Convert functional test data

The set of data used for the MIP Convert functional tests is a bit disorganised. As we add new functional tests it would be good to create a new structure to store the minimum amount of data in (to speed the tests).

Initially CMIP7 data should be held separately and we can then migrate CMIP6 data over.

Consider
* should we hold each data separately for each test?
* should we have a mechanism for documenting where the data came from (e.g. a script to pull data from MASS)
* range of model runs we pull data from