Retrain and save fresh recommendations on a schedule #71

Blocked by: #70

Problem

New events only change the rates the manager sees after the model retrains and new recommendations are saved, and both only happen when someone triggers them by hand.

Goal

After each data refresh, the model retrains when needed and fresh recommendations for the next 30 nights are saved, so rates reflect new events within hours.

Scope

• Add retraining and saving recommendations to the schedule, after the events and weather refresh.
• Decide the order and timing: refresh, then retrain, then save recommendations. Retrain on every run or less often, depending on how long it takes.
• The same failure visibility as the daily refresh.

Considerations

• How long and how much a retrain costs on the hotel's real data decides how often it can run.
• Training while a refresh is still running would use partial data; wait for the refresh jobs to finish first.
• Once manager feedback exists, the next scheduled retrain should include it.

Acceptance criteria

• Without anyone calling the API, a new event leads to updated recommendations within the agreed time.
• Each scheduled run saves recommendations for the next 30 nights.
• Failed retrains or saves are visible to the team.