Bug: TUNIC: 'all state' constructed in stage_extend_hint_information collects most items twice

[Mysteryem]

What happened?

multiworld.get_all_state() cannot be used to produce a valid 'all state' after Core AP's main fill has been run because get_all_state() collects all items from multiworld.itempool and all reachable placed items, which typically results in collecting every item into the 'all state' twice.

With default arguments, get_all_state() also collects items from each worlds' get_pre_fill_items(), which would generally be invalid to do after the pre_fill step because those items should be placed after the pre_fill step, so collecting items from each world's get_pre_fill_items() would often result in collecting those items a second time.

Archipelago/worlds/tunic/__init__.py

Lines 676 to 683 in 6875f9d

	def stage_extend_hint_information(cls, multiworld: MultiWorld, hint_data: dict[int, dict[int, str]]) -> None:
	tunic_er_worlds: list[TunicWorld] = [world for world in multiworld.get_game_worlds("TUNIC")
	if world.options.entrance_rando]
	if not tunic_er_worlds:
	return
	hint_data.update({world.player: {} for world in tunic_er_worlds})
	all_state = multiworld.get_all_state()

multiworld.get_all_state():

Archipelago/BaseClasses.py

Lines 452 to 464 in 6875f9d

	ret = CollectionState(self, allow_partial_entrances)
	for item in self.itempool:
	self.worlds[item.player].collect(ret, item)
	if collect_pre_fill_items:
	for player in self.player_ids:
	subworld = self.worlds[player]
	for item in subworld.get_pre_fill_items():
	subworld.collect(ret, item)
	if perform_sweep:
	ret.sweep_for_advancements()
	return ret

What were the expected results?

The 'all state' that is constructed should only collect each Item instance at most once.

Software

Local generation