Fix Dargo's sacrifice cost

[Callumvl]

Hopefully addresses #12349 + #11443
as well as part of #7750

The previous implementation would suddenly add large discounts on casting a creature/artifact, didn't seem to handle rollbacks, among other issues.

fix #12349, fix #11443

[xenohedron]

this looks fine, using common classes is good, but I am confused as to what the problem actually was

[Callumvl]

@xenohedron I am still somewhat confused myself on the problem. But happy to provide what I did work out.

The main issue is that Dargo was sometimes being discounted to just R when there had not been a sacrifice that turn (this was an AI match, not sure if relevant). I added some temporary debug messages to where the sac map is incremented and noticed that i seemed to get SACRIFICED_PERMANENT events when i was not expecting (e.g. on casting something), causing the value in the sac map to be incorrect.

To replicate:

1. Start an AI match with Dargo as one of the commanders.
2. Add 5 Lotus Petal to hand.
3. Cast 1 Lotus Petal.
4. Dargo appears as castable for R (should be 6R)
5. Cast Dargo using the Lotus Petal (should have cost 4R with a -2 reduction from the Lotus Petal sacrifice)

While looking into this I also noticed a secondary issue, using the rollback function does not reset the sacrifices for that turn. So on a turn where Dargo costs 6R, if the cost is reduced via sacrifices to R, then the turn is rolled back, he will still cost R without having to sacrifice anything.

Hope that clarifies things somewhat,
Cheers!

[xenohedron]

Interesting. The AI does simulate the game state, which effectively involves firing various events and rolling back, and those simulations do get caught by print statements. You can wrap in if (!game.isSimulation()) { } if needed. But those simulated states should not be getting picked up by the watcher, not sure what the cause might be.

So you confirm that you can no longer reproduce those bugs with the code in this PR?

[Callumvl]

@xenohedron Thanks for the suggestion. I believe I may have found the culprit, all of the increases were related to simulated events, the problem was that the sacMap was static,. After removing that part of the declaration, the bugs are also fixed.

Yes, i can confirm :)

[JayDi85]

@xenohedron looks like there are many buggy places with same code style (wrong usage of static fields). Created related issue: #14249

[xenohedron]

Oh good catch, that certainly explains it