Attack chain, initial setup. (pull *immediately* for *any* TM issues) (#26834)

* refactor: Attack chain, initial setup.

* migrate curtain to make dreamchecker happy

* update thurible

* don't call attacked_by separately for legacy attack chain

* remove duplicate proc

* condense similar code, put allowances for legacy code in new procs

* update docs, include diagram source

* add comment on how to update diagram

* fix admonition

* mindflayer updates

* remove commented out code

* clarify all steps

* after_attack should be overridable

* whoops

* retrofit recent changes

* duh, can't restrict this yet because of tool_acts

* i hate ore bags with the fire of a thousand suns

* return correct value for object attack logic

* Various cleanups.

We don't want to attempt to pull stuff out of `/obj/item/attackby`,
because those pieces are part of the related objects' migrations, not
`/obj/item` itself. Attempting to do this causes knockon effects where
things expected to call e.g. `/obj/item/storage/attackby` in the call
chain were not ferried over to the new item interaction code, because
the related objects hadn't actually been migrated over yet.

I've used refactoring /obj/vehicle as the example for migrating
`attackby` methods instead.

* simplify some argument names

* fuck it

* make it do the thing

* Rename CI module call

* Prove that CI works

* improve test output

* aaand fix it again

* fix curtain tool interactions

* fix compile error

* fix compile error

* Better docs, introduce migration plan tool.

code/modules/admin/topic.dm

@@ -2011,7 +2011,7 @@
 					possible_guardians -= "Random"
 					scarab.possible_guardians = list()
 					scarab.possible_guardians += typechoice
-				scarab.attack_self(H)
+				scarab.attack_self__legacy__attackchain(H)
 				spawn(700)
 					qdel(scarab)
 				logmsg = "scarab guardian."