use chunks.archive.d for borg info repo::archive?
check if we can speed up using the single-archive hashindex cached there.
note: for my own ~500GB repo and a ~130GB archive, borg info repo::archive took ~6 minutes.
This was actually already implemented once in #2764
Ah, right, we wanted to cherry-pick that part from there... - can you do it?
It heavily depended on the whole patch.
part of this was re-implemented in #4391.
the remaining slow path is for computing the unique_csize (and fetching / building an archive index from remote for this).
i removed all the csize stuff. the unique size computation is still expensive.