AI must support multiple targets on resolve

[JayDi85]

From #12039 (comment)

BarterInBloodTest uses AI targeting instead direct test commands and it fails on "while" remove. Looks like AI do not support multiple targets in choice dialogs (on resolve) and it works fine for cast/activate only (cause it uses game simulations for it).

Good code from human:

Bad code from computer:

It's a more global problem with AI, not related to PR. Possible bug: ability require to select a multiple targets for sacrifice, but AI selects only 1;

TODO:

• add AI tests for "choose multiple targets on resolve" use cases -- BarterInBloodTest and SacrificeAllEffect is good starting point;
• search and replace all while cycle with Target::isChosen usage by single "choose" call (all multiple choices must be processed inside a choose dialog);

[JayDi85]

Fix example: 39872e2

[JayDi85]

Looks like TestPlayer can have same problems -- if you setup wrong choices in VonasHungerTest->test_Ascend_Normal (e.g. choose 1 creature instead 3) then it pass it with 1 sacrificed creature and without any errors.

Test code must raise errors on wrong targets setup. Need research -- is it possible to fix. Yes, it's possible to fix, see example with c82511f