Tests: Don't Modify Rules After set_rules

[Exempt-Medic]

What is this fixing or adding?

Adds a test that worlds don't modify Location.access_rule or Entrance.access_rule after set_rules. Also clarified this in the world API doc.

It's very barebones, there's probably a better way to format and perform the test.

Ocarina of Time and Pokemon Red and Blue are exempted from the test because they are exempt from the test_location_creation_steps test (they remove locations in generate_basic and pre_fill)

How was this tested?

The test, which TLoZ fails

• Bug: Worlds Applying Rules After set_rules #4563

[NewSoupVi]

Couple of considerations.

1. I think it's fine to modify rules after set_rules. I think a fully functional set should be present after set_rules, but I 100% think that changing them again afterwards has some valid use cases, like banning an item from a certain area for "fun" reasons by using an overly restrictive rule, but then switching back to the "actual" logic for spoiler log output.

2. I think if entrances are allowed to stay unconnected until connect_entrances, they should be allowed to change rules all the way up until then as well. I can think of some scenarios where an entrance rando implementation using GER might want to change rules depending on which entrances were connected via a custom on_connect.

That said, I don't fully disagree with the need for a test or further documentation. I definitely think that setting your first ever rules in generate_basic is bad, for example. I wish I had a better counter proposal, but I'm coming up blank right now.