[Split Tracking] Modify submission pipeline to write split records and adjust group leaderboard

[joelhalen]

Overview

Part of #28 — Split GP Tracking feature.

When a drop is submitted with players_included for a group that has split_gp_tracking enabled, the pipeline must:

1. Persist split credit records to drop_splits.
2. Adjust the group leaderboard in Redis so each participant (including receiver) receives an equal share of the GP value.

Work Required

File: data/submissions/drop.py — drop_processor()

Insert after the existing group config loading block (~line 320, after player_groups is resolved and group configs are fetched):

For each group in player_groups:

1. Check if split_gp_tracking config is "1" for this group.
2. If enabled and players_included is non-empty:
   a. Resolve each name in players_included to a Player row (query by player_name).
   b. Filter to players who are valid members of this group — reuse check_split_capability() from data/submissions/point_awards.py.
   c. Build participants = [receiver] + [valid split players].
   d. split_value = (drop.value * drop.quantity) // len(participants).
   e. For each non-receiver participant:
   • Create a DropSplit(drop_id=drop.drop_id, player_id=p.player_id, group_id=group.group_id, split_value=split_value) and flush.
   • Call redis_updates.add_split_credit(p.player_id, split_value, partition, group.group_id, world_type).
     f. Adjust the receiver's group leaderboard score: call redis_updates.add_split_credit(receiver_id, split_value - full_value, partition, group.group_id, world_type). This is a negative delta that brings the receiver from full_value down to split_value on the group board only — the global leaderboard and player:*:total_loot keys are untouched.

Key Constraints

• The global individual leaderboard and player:{id}:{partition}:total_loot Redis keys are never modified here — only leaderboard:{partition}:group:{group_id} sorted sets.
• Groups without split_gp_tracking are unaffected; no DropSplit rows are written.
• players_included is already normalised at line 110 via _normalize_incoming_players(); use that result directly.