CLI/git: current cache contents of index

Rather than shelling out once per message to get the list of files
corresponding to tags, it is much faster (although potentially a bit
memory intensive) to read them all at once.