I really don’t think you’re considering all the edge cases in such a scenario.

Most devs have problems writing decent queries (in some situations), now you want to introduce writing the query execution plan into the mix ?

No one is taking away the existing automatic planning that works well 95% of the time. You're welcome to continue using that.

The worst thing that could possibly happen is that you give it bad advice leading to slow queries, and then the obvious first step to fixing that is to drop the manual advice and see whether the automatic planner handles it better.

It baffles me that PostgreSQL, which is so deeply customisable in almost every other way, resisted this form of customisability for so long. This is great news.