Flickr-download

Adds metadata store option for `--download-user` too. Unclear how that was missed previously :)

* fixes JSON metadata downloads when not being the owner of the photo
* allows newer versions of python-dateutil and upgrades to latest

Upgrades to 0.7.7 of [flickr-api](https://github.com/alexis-mignon/python-flickr-api) which adds:
* timeouts to API calls
* If no size is specified, prefer `Original` over other equal largest sizes as suggested by zanna-37
* improves the help documentation by zanna-37

Managed to break the increment feature of file naming in the previous build. This should fix it

Fixed the underlying Flickr API to handle video file sizes better, and removed some old cruft code for video download handling.

This also delegates the responsibility of file extensions to the underlying API.

This release has a few speed improvements that especially will speed up things when downloading large sets, and having to restart because the Flickr API inevitably errors out at some point:

* You can now add the `--metadata` flag to have the script keep track of which files have been downloaded in a way that eliminates the API call to Flickr entirely for an already downloaded image
* The `--cache` flag now also caches the set list
* One extraneous API call per image download removed

And then a bunch of fixes:
* Actually iterate through all photosets, photos, etc (and not only the first 500)
* Handle error where photo date is too old for the underlying OS to handle

Behind the scenes, there are now a few more tests and everything is type annotated.