Switch get_node_details to always use cache when available...

[kollivier]

…and update cache values async.

Description

Since get_node_details is often called on the root node and can take significant time, switch to always preferring cache results when available.

Issue Addressed (if applicable)

A whole bunch of Sentry reports about get_topic_details timing out ;-)

Steps to Test

• Add a node or nodes to a channel.
• Check details in channels page, they should be the old values
• Check again a short time later - they should be updated.

Implementation Notes (optional)

At a high level, how did you implement this?

Much of this PR is shuffling code around, as in order to call get_node_details as a task I needed to move it out of views/nodes.py in order to avoid a circular import dependency between views/nodes.py and tasks.py. For similar reasons, I needed to move the thumbnail handling code outside of views/nodes.py as well.

The best place it seemed to put these methods was within the models themselves, so I added ContentNode.get_details and ContentNode.get_thumbnail. I also switched the get_details function to call Channel.get_thumbnail() instead of using the get_channel_thumbnail function in views/nodes.py, which could then be removed.

Lastly, I added tests ensuring that get_details() still returns the relevant information, that calling it from the get_topic_details endpoint updates the cache when needed, and that the get_thumbnail model methods return the correct thumbnail info.

Checklist

Delete any items that don't apply

• Is the code clean and well-commented?
• Have the changes been added to the CHANGELOG?
• Are there tests for this change?

Reviewers

If you are looking to assign a reviewer, here are some options:

• Jordan jayoshih (full stack)
• Aron aronasorman (back end, devops)
• Micah micahscopes (full stack)
• Kevin kollivier (back end)
• Ivan ivanistheone (Ricecooker)
• Richard rtibbles (full stack, Kolibri)
• Radina @radinamatic (documentation)

[jayoshih]

Looks great! Thanks for getting this out so quickly! Some super minor stuff, but otherwise should be good to merge

[jayoshih]

I did implement a function to handle this without relying on ast.literal_eval. Do you mind switching to this?

[kollivier]

Sure, will do! :)

[rtibbles]

Note that for channel thumbnails I migrated the thumbnail_encoding field to a JSON field to avoid this malarkey: #1140

The data migrations that I made to facilitate this should transparently transfer to the content node models also.

[kollivier]

Yes, I was thinking about doing this, but then I thought that if we go the route of migrating to all files for ContentNode thumbnails, we can just create a migration to remove thumbnail_encoding entirely on ContentNode. :)

[rtibbles]

Not in the scope of this PR, but I note that we are currently using the database cache backend: https://github.com/learningequality/studio/blob/develop/contentcuration/contentcuration/settings.py#L94

I imagine this may give us a slight performance boost because it will always be one query, than potentially many - but this will still be much slower than using an in memory backend (I believe @kollivier had previously suggested https://github.com/sebleier/django-redis-cache for such purposes - he's a smart guy, so I'd definitely listen to him!)