[back][ext] feat: recommendations API can now return random entities

[GresilleSiffle]

Description

This PR adds two features:

(1) A new API polls/{name}/recommendations/random/ that returns random recommendations, designed to be more efficient than the regular recommendations route.

(2) The browser extension now uses this new API to improve the loading time of recommendations and their diversity.

context

This PR aims to reduce the number of latency alerts triggered by the API recommendations, by providing API clients, such as the browser extension, an alternative and efficient way to request random recommended entities.

Those latency alerts are currently triggered by the browser extension, when fetching the top 250 of all time recommendations to perform a random selection. Thanks to the new API, the extension doesn't need to fetch a big amount of entities to display a diverse bundle of entities.

Note that the new API is more efficient because it uses a simple SQL query that provides less features than the regular recommendations route:

• no search by text
• no search by weighted criteria score
• not possible to return unsafe recommendations

ℹ️ consequences and important changes

This PR significantly changes the strategy used by browser extension to display diverse entities.

(a) Before, only the top 250 of all time recommendations could be displayed. Now all recommended videos have a chance to be displayed, even those with a Tournesol score of 20.1, like on the web site.

(b) The number of cached recommendation results requested by the extension will increase from 2 every 10 minutes (for a given set of languages) to ~ 40, up to a lot more depending on how many time the users refresh their recommendations with the extension refresh button. This may impact our infrastructure.

to-do

back end - /recommendations/

• remove the random logic

back end - /recommendations/random/

• add a /recommendations/random/ view
• create a new serializer
  • and the random parameter in the metadata?
• document the view
• test the view

extension

• request random recommendations
  • should the recommendation be in the top X ?
• randomly initialize the query parameter random for each instance of TournesolRecommendations
• make the query parameter random vary for each refresh
• increase the limit parameters for more randomness
  • actually 5 for old videos
  • actually 7 for new videos
• connect the extension to the new route

Checklist

• I added the related issue(s) id in the related issues section (if any)
  • if not, delete the related issues section
• I described my changes and my decisions in the PR description
• I read the development guidelines of the CONTRIBUTING.md
• The tests pass and have been updated if relevant
• The code quality check pass

[GresilleSiffle]

I've been able to significantly improve the performance by removing the total_score from the SQL query, and the related left outer join tournesol_entitycriteriascore. I think this should our first optimization.

[GresilleSiffle]

The check run end-to-end tests / e2e-browser-extension (pull_request) is failing because the browser extension uses the route that comes with this PR (not deployed in production yet).

[lfaucon]

They don't necessarily need to be the same. The backend can assume 10 when nothing is specified & the frontend can initialize a user's search by default with a value of 50.

I vaguely remember discussing with @amatissart to make it 0 by default. I don't remember well what makes a criteria included or not in the scores used for ranking.