This is closely related to MDL-59067

Community of inquiry indicators' calculations are based on:

• The % of activities that were expected to be completed during a specific period that the user interacted with
• How "deep" was this interaction

As we calculate indicators once the target is ready (the end of the course for students at risk of dropping out courses) we need to guess the activities that were available at that time. MDL-59067 could solve this but sites will still have a lot of finished courses we want to use to train machine learning algorithms. This issue is about improving the current "guess" system that can be used as a fallback in case the system can not find indicators calculations stored as part of MDL-59067.

Some comments copied from code TODOs:

• In completed_by: Use course_modules_completion's timemodified + COMPLETION_COMPLETE* to discard activities that have already been completed.
• In availability_completely_by: Would be nice to expand \availability_date\condition API for this calling a save that does not save is weird.