Instead of hard-coding a check for pinned items when storing the first
item, allows the creater of NodeIterator to specify how the comparison
should be made.
This also allows storing the newest post (whatever it's status is),
instead of simply storing the newst non-pinned post, which prevents
redownloads in case a user adds only a pinned post between runs.
The most recent post is cached in NodeIterator (and
saved to the disk), and its timestamp is used, instead of the
timestamp instaloader was run.
This way, even in later resuming runs the timestamp stored is the same
that would have been stored if the first run.
Fixes#1206.
Using itertools.takewhile() on a NodeIterator returns a plain Iterator,
and so it's not resumable.
The strategy has been altered to pass an extra argument to
posts_download_loop, a lambda that is evaluated for each post, and
causes the loop to stop when it returns false.
While this isn't really an error, it requires user action, so the message
is printed to stderr even with --quiet, and repeated at the end when that
option is not enabled.
Resolves#1159.
Adds the --latest-stamps command line option, pointing to a file where
the latest time each profile was scraped. On the next run, only posts
newer than that time are downloaded.
Fixes#1122.
