Fix shutdown crash when flushing actions for unregistered players

[luozaixuan]

Ledger (1.3.18, 1.20.1 Fabric) crashed 100% of the time on server stop when the action queue had a backlog (tens of thousands to hundreds of thousands of entries): flushing inserted into the players table and hit SQLITE_CONSTRAINT_NOTNULL (players.player_name is NOT NULL with no default), the error was mistaken for a transient failure and retried 3 times with backoff, stalling shutdown for ~5 minutes before timing out and losing the whole backlog.

Root cause: insertActions called getOrCreatePlayerId(it.id); that helper reused getOrCreateObjectId, which is designed for single-key lookup tables (Sources/Worlds/ObjectResourceLocations) and, on a miss, inserts only the unique key column. The Players table has a second mandatory column, player_name, so the generated INSERT omitted it and violated the constraint. Player rows are normally created asynchronously by onJoin's logPlayer, so any action produced before that write landed (or by entities/Create fake players with no registered join) hit a missing player row when shutdown force-drained the queue.

Fix: getOrCreatePlayerId now takes the full GameProfile and, on a cache/DB miss, inserts a complete row including player_name (falling back to an unknown_ + uuid-prefix name within the 16-char limit when the profile has no name), keeping both playerKeys and playernameKeys caches in sync.

No schema or migration changes are needed: the NOT NULL constraint guarantees no dirty rows exist in existing databases.

[luozaixuan]

this fix only for 1.20.1-backport,because Ledger with Create will have this error