Don't rely on the HTMLFile model to index and re-index files

[stsewd]

So, we don't need these objects in the db to re-index, since we re-read all their data from storage to index/re-index.
The only benefit is to track changed files, but that is going to be removed in #7874.

Our current flow is:

• Create html objects from all html files in storage
• Create intersphinx data for existing html files (here we use the html file objects to check that, but it can be replaced by keeping files in memory)
• Sync to ES
• Delete html files from previous builds (afte this, the html objects aren't used for anything else other thant the re-index management command)

Benefits include not having that data in the db (more spaaaace!), and removing the django-elasticsearch-dsl package (fewer dependencies!).

Work to do:

• Use a structure/class instead of an ORM model to pass around the data in the indexing step.
• Change our re-indexing management command to not depend on the objects from the db, but do the re-index directly.

[humitos]

This was implemented in #10696

removing the django-elasticsearch-dsl package (fewer dependencies!).

This is going to be done in #10730