Improve how AI chooses damaging moves

[DizzyEggg]

Fixes #3130
Fixes #3153

The general flow should be like this(assuming there are no special effects which would otherwise be more preferable):
if can faint -> choose the move without side effects(e.g. recoil, recharge, loading turn), and with better accuracy(e.g. Tackle over Mega Punch if both are 1hko)
if can't faint -> choose the move which takes the least no of hits to KO, but only if the other move doesn't have any advantages(otherwise it's a toss-up and depends on the battle conditions).

Also fixed Taunt's score to be minus 10 and improved handling of Solar Beam.

Before:

After:

[AlexOn1ine]

What do you think about removing this switch block?

        switch (AI_DATA->effectiveness[battlerAtk][battlerDef][AI_THINKING_STRUCT->movesetIndex])
        {
        case AI_EFFECTIVENESS_x8:
            score += 8;
            break;
        case AI_EFFECTIVENESS_x4:
            score += 4;
            break;
        case AI_EFFECTIVENESS_x2:
            if (AI_RandLessThan(176))
                score += 2;
            else
                score++;
            break;
        }

Your PR takes care of a lot of edge cases and this block looks unnecessary with it. You would just get score increases for the sake of increasing it.

Also how valuable are these checks?

        if (gBattleMoves[move].highCritRatio)
            score += 2; // crit makes it more likely to make them faint

    if (!WillAIStrikeFirst() && CanTargetFaintAi(battlerDef, battlerAtk))
    { // AI_TryToFaint_Danger
        if (GetMoveDamageResult(move) != MOVE_POWER_BEST)
            score--;
        else
            score++;
    }

Since you made the AI pick the best damaging move in AI_CheckViability,it seems redundant to check again in AI_TryToFaint

    if (gBattleMoves[move].power)
        score += AI_CompareDamagingMoves(battlerAtk, battlerDef, AI_THINKING_STRUCT->movesetIndex);

/Edit: resolved in discord

[AlexOn1ine]

I did some testing and my suggestion generally would be to only decrease the score when compared to a better move.
The problem with increases is that it inflates the score since each move is compared against each other which might render status moves unusable.
Let's say we have the base

Good move, score 100
Bad Move, score 100
Bad Move, score 100
Bad Move, score 100

If the comparison from this PR are applied it ends up like this

Good move, score 103
Bad Move, score 99
Bad Move, score 99
Bad Move, score 99

My suggestion would be to decrease the bad moves only

Good move, score 100
Bad Move, score 99 
Bad Move, score 99
Bad Move, score 99

Same result but you don't inflate the score

The problem gets worse if there are 2 different bad moves present, like a less accurate move and a recoil move.

/Edit: I just noticed the decreases are somewhat inconsistent in the PR. Sometimes it increases and decreases, sometimes it only increases. Like I said my suggestion would be to only decrease but regardless I would make it consistent across the board.

[DizzyEggg]

Despite the issue you Alex mentioned I still think this PR does more good than evil. Would it be fine to merge it to upcoming and then improve it? The thing is even if it's not perfect with this PR, it is still worse without it. Also, I hope to have AI tests soon, so it'll be easier to catch these issues.

[DizzyEggg]

@AsparagusEduardo Should be ready to merge.