directory content is only implicitly known

when storing a directory, only the directory name and some metadata is stored into the repo, but not the names it contains.
- each file is stored with full path, thus using that, they are restored into the correct directory.
- but, at extraction time, we can not process a list of directory contents and KNOW when we are finished with it. we just encounter the directory member names while extracting files, but we never know when we are finished with a directory.
- if operating strictly sequentially, we can deduct when we are finished with a directory because we know the traversal mechanism that created the archive. but if that strict sequential processing is not available (e.g. due to multithreaded processing in future), we can not set the directory mtime until the very end of the extract operation, because there could be always another directory member coming...