Update winners picking logic (#3073)

# Description
In a discussion with the solver team, several different algorithms were
proposed to pick winners.

https://www.notion.so/cownation/Rewards-for-multiple-winners-1218da5f04ca80f2a59ed8f0174b3a69?pvs=4#1228da5f04ca8050ac31e965b120709d

> We choose winners iteratively:
> 
> 1. Choose solutions with largest score from remaining solutions as
next winner.
> 2. Choose largest score of solutions with overlap to winner as
reference score. The reward is winning score - reference score, the
penalty is reference score.
> 3. Remove all remaining solutions which have token pairs in common
with the winner. Whenever a solution is filtered in this procedure,
additionally use all token pairs of that filtered solution for filtering
remaining solutions.
> 4. Repeat.

Currently, the best option is "second price with aggressive filtering".
This PR implements this logic.

# Changes
<!-- List of detailed changes (how the change is accomplished) -->

- [ ] Instead of adding only winner's swapped tokens to already swapped
tokens list, add every solution's swapped tokens, regardless if they are
a winner or not.

crates/autopilot/src/run_loop.rs

@@ -564,8 +564,8 @@ impl RunLoop {
 
         // Winners are selected one by one, starting from the best solution,
         // until `max_winners_per_auction` are selected. The solution is a winner
-        // if it swaps tokens that are not yet swapped by any other already
-        // selected winner.
+        // if it swaps tokens that are not yet swapped by any previously processed
+        // solution.
         let mut already_swapped_tokens = HashSet::new();
         let mut winners = 0;
         let solutions = solutions
@@ -581,10 +581,8 @@ impl RunLoop {
                 let is_winner = swapped_tokens.is_disjoint(&already_swapped_tokens)
                     && winners < self.config.max_winners_per_auction;
 
-                if is_winner {
-                    already_swapped_tokens.extend(swapped_tokens);
-                    winners += 1;
-                }
+                already_swapped_tokens.extend(swapped_tokens);
+                winners += usize::from(is_winner);
 
                 participant.rank(is_winner)
             })