v0.11.1 — Fix /save-all flush hang during operation

Critical hotfix for a regression: running /save-all flush while the server is live could spin the main thread in an endless loop, tripping the 60-second ServerWatchdog kill and freezing the server. This affects any path that issues /save-all flush during operation — an admin running it by hand, backup mods snapshotting before a save, or a scheduled save task.

The problem

Vanilla saveAllChunks(flush=true) keeps re-scanning every chunk "as long as any chunk got written this round," stopping only once a full pass writes nothing. To advance vanilla's save cooldown, BAS reports already-clean chunks as "saved"; during operation that report keeps the loop's condition permanently true, so the main thread never leaves the save loop and the watchdog declares a hang after 60 seconds.

Only the operational /save-all flush reaches this path. Shutdown flush is unaffected (it runs a different synchronous drain), and periodic autosave is unaffected.

The fix

Operational /save-all flush no longer hands control to vanilla's synchronous loop. It now uses the same async enqueue path as periodic autosave — dirty chunks go to the background serialization workers and the command returns in well under a second. Shutdown flush keeps its existing synchronous drain semantics unchanged.

Upgrading

Drop-in jar replacement, no config changes. Every 0.11.0 user should upgrade: if anything on the server can issue /save-all flush, 0.11.0 can be watchdog-killed after a few hours of uptime.

Fixes #5, #6