Currently we reread the file for every new link we download.
In cases where user runs youtube_dl to keep their library
up-to-date, this is very inefficient, as most of the links
are already downloaded.
Archive class works half way between a database and an append
log. It's an append log that also keeps a hash set in memory
for fast checking existing links.
A compatibility function was added for reading file's last
modification date
Tests have been added
