[Split Tracking] Add `drop_splits` database table and ORM model

[joelhalen]

Overview

Part of #28 — Split GP Tracking feature.

A new table is needed to persist which players received split GP credit for a given drop within a given group. This is the source-of-truth that the Redis layer uses when force-rebuilding leaderboards.

Work Required

New file: db/models/drop_split.py

class DropSplit(Base):
    __tablename__ = 'drop_splits'
    id          = Column(Integer, primary_key=True, autoincrement=True)
    drop_id     = Column(Integer, ForeignKey('drops.drop_id'), nullable=False, index=True)
    player_id   = Column(Integer, ForeignKey('players.player_id'), nullable=False, index=True)
    group_id    = Column(Integer, ForeignKey('groups.group_id'), nullable=False, index=True)
    split_value = Column(Integer, nullable=False)   # GP credited to this participant
    date_added  = Column(DateTime, default=func.now())

• Add DropSplit import to db/models/__init__.py (follow the existing import pattern).
• Add a CREATE TABLE IF NOT EXISTS drop_splits (...) statement to the DB initialisation script / Alembic migration so the table is created on startup.

Notes

• player_id here is a split participant (non-receiver). The drop receiver is identified via drops.player_id.
• split_value stores the per-participant GP credit (i.e. total_drop_value / num_participants).
• Only rows for groups that have split_gp_tracking = 1 will ever be written here.