`get_labeled_inventory` needs better performance for large databases

[cmutel]

get_activity(lca.dicts.biosphere.reversed[i]).as_dict() will be slow (1 query per object). Should be 1 database query in total.

[BenPortner]

Perhaps, one could re-use bw2data.backends.iotable.storage.Storage for this? It might be worthwhile investing in a central cache and handler, which can be used to add activity metadata to all kinds of arrays and matrices, be it exchanges, inventory, biosphere_matrix or technosphere_matrix.