Better youtube_dl requests mgmt

[rgaudin]

The new structure from #119 PR accidentally strengthen something wrong in our requests management. Every youtube_dl call leads to a request to the video webpage from which youtube_dl extracts the information it needs to then

• download the video file(s)
• download the thumbnail(s)
• download the subtitle(s)
  As this webpage is subject to the quota, this is not very efficient as we'd call it 3 times in case we have nothing for that video in S3.

We should also probably save those subtitles in S3 although those can change over time. As for thumbnails, I don't think this matters much until users raise concern about subtitles being outdated (then we'd expire them after some time).

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be now be reviewed manually. Thank you for your contributions.

[benoit74]

Need to be done together with #177

And caching S3 subtitles is now tracked as #277