fix: remove sentinel-player dependency in teams-left

[mateu]

What

Remove hardcoded player = 24 dependency from teams-left leaderboard metrics.

Why

count_player_teams_left and count_player_final4_teams_left inferred remaining games from one sentinel player's picks, which silently breaks when that player row is missing or incomplete.

How

• Reworked both SQL CTEs to derive remaining games from game rows not present in played games (player = 1) instead of a sentinel player's picks.
• Kept existing loss-elimination logic intact.
• Added safe max_left handling for empty non-admin player sets.
• Updated regression coverage to assert sentinel-free behavior and added explicit checks for both teams-left methods.

Testing

• script/test-env.sh prove -lv t/model_round_based_final4_counts.t
• script/test-env.sh prove -lr t

---

Quality Report

Changes: 2 files changed, 30 insertions(+), 25 deletions(-)

Code scan: clean

Tests: skipped

Branch hygiene: clean

Generated by Kōan post-mission quality pipeline

[Copilot]

Pull request overview

This PR removes the hardcoded dependency on a sentinel player = 24 for computing “teams left” leaderboard metrics, making the calculation resilient when that player row/picks are missing.

Changes:

• Reworked count_player_teams_left and count_player_final4_teams_left to derive games_remaining from game rows not present in “played games” (player 1 picks), rather than from a sentinel player’s picks.
• Added safer max_left handling when there are no non-admin players in the computed result set.
• Updated/extended regression tests to ensure sentinel-free behavior for both teams-left methods.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
lib/Bracket/Model/DBIC.pm	Removes sentinel-player CTE dependency and adds empty-set-safe max_left computation for both teams-left queries.
t/model_round_based_final4_counts.t	Updates fixture to omit player 24 and adds assertions covering sentinel-free behavior.

[Copilot]

The new empty-non-admin-player branch for $max_left (returning 0 when there are no players other than player 1 in the result map) isn’t covered by tests. Add a regression case where only the admin/perfect player exists (or where no non-admin players have remaining teams) and assert $max_left is 0, to prevent reintroducing the previous max-on-empty behavior.

[Copilot]

Similar to count_player_teams_left: the new $max_left fallback to 0 when there are no non-admin players in the computed map is not exercised by the current test. Add a regression test for count_player_final4_teams_left where no non-admin players are present (or none have remaining final4 teams) and assert $max_left is 0.

[mateu]

@copilot apply changes based on the comments in this thread

[Copilot]

@copilot apply changes based on the comments in this thread

Done in adeaa51. Added two regression tests right after the UDFs are registered and before any non-admin player picks are created — at that point the result map is empty, @players is an empty list, and both count_player_teams_left and count_player_final4_teams_left must return $max_left = 0, directly exercising the new fallback branch in each function.