The current PGA index format does not allow to understand under which references a given URL is written. For example, tensorflow/tensorflow belongs to 2 siva files, the first has two heads and the second - tens. I need to have the full references collection in the second CSV column, e.g.

"c19e4a1b8c7f458fa4d6b0978e2a14ef8c2a2ff2.siva[refs/heads/<uuid>,refs/whatever/<uuid>],f41959ccb2d9d4c722fe8fc3351401d53bcf4900.siva[refs/heads/<uuid>,...]"