Improving ma_qa_metric_local_pairwise description

[aozalevsky]

right now ma_qa_metric_local_pairwise doesn't have any details about how complete the data should be. For instance, a PAE matrix is supposed to be a symmetric square matrix. Thus, only an upper triangular matrix should be enough. I guess the initial definition was intentionally generic, but maybe we can extend the category description with some case-specific (like PAE) details. To give some guidance for software developers.

This came up as a part of the discussion in chaidiscovery/chai-lab#52

[aozalevsky]

Re-created in ihmwg/ModelCIF#19, so closing this one.