This allows restricting each CA individually to only a subset of the units. It also means that several instances of the same CA can be used in the same AI configuration, in order to, for example, force the order in which units are dealt with, or to apply different settings to different units.
Also backtracked the removal of wesnoth.get_recall_units, which is now available as wesnoth.units.find_on_recall - it's just more convenient than passing x=recall to wesnoth.units.find
Now that all the AIs use external CAs, there is no need to use the persistent 'data' variable any more, unless information is to be exchanged between different CAs or is supposed to be persistent across save/load cycles.
(cherry-picked from commit 3bfd59f28ba7f70a6ac32782e98cba9ca6c2a44a)
These used to be accessible only through unit.__cfg or wesnoth.unit_types. The Fast Micro AI is not included here as it requires a larger clean-up.
(cherry-picked from commit 8af988c6972abe75e985d0e43bd8186fe4826cb6)
Otherwise it is too easy to exploit this by baiting the AI leader to
run into a trap, and it is too difficult for what it's worth to make
sure the attack location is safe. If only the side leader is left to do
such an attack, it's more often than not not a good idea to do so
anyway.
Because the attack simulation during CA evaluation is done with enemy
XP set to zero, the AI previously sometimes chose a different weapon
during execution than what was used for evaluation. This resulted, for
example, in archers sometimes attacking an enemy without ranged attacks
with their melee attack when aggression=1.
This CA performs attacks on enemy units so close to leveling that the
default AI's combat CA would not attack them (with some exceptions).
This is meant to keep players from being able to exploit this known
weakness of the default AI.
