Fix AI implementation of Toxic Deluge and Flowstone Slide

[liamiak]

Both carried AI:RemoveDeck:All because the AI could not cast them at all. Their X is announced as a cost and nothing ever chose it — PumpAll has an AI class, so AiController's generic X handling is skipped, and Toxic Deluge's X lives in a PayLife<X> cost that path would not have seen anyway. With X unset both pump amounts read 0, so every sweep evaluated as -0/-0 and was declined. Toxic Deluge was also missing IsCurse$ True, so PumpAllAi never reached its sweeper branch and instead asked whether to pump the AI's own creatures.

PumpAllAi now picks X for a -X/-X sweep when the ability announces its own X, scoring by what each value destroys. Candidates are the toughnesses actually on the battlefield rather than every point up to the maximum, so the AI does not walk a 40 point life total one step at a time.

Rather than reimplementing combat reasoning, the two questions it needs are handed to the code that already answers them. Whether the swing simply wins goes to the attack planner, which knows about fog, evasion and cantWin. Whether the AI lives through the counterattack goes to predictNextCombatsRemainingLife, which models the blocks it will make — and is only asked when the sweep is not lethal and a cheap pessimistic bound says the survivors could get there, so it is paid for exactly where it changes the answer.

Each needed one small addition, in its own commit: a counterpart to removeBlocker for our own creatures, and a way to tell combat prediction which attackers a wipe removes.

Best read commit by commit — a shared predicate, the two enabling additions, the fix, then tests.