For each player on each tick, enter block triggers are invoked, and these create loot contexts that are promptly thrown away since the trigger doesn't pass the predicate
To avoid this, we now lazily create the LootContext if the criterion passes the predicate AND if any of the listener triggers require a loot context instance
On each tick, Paper runs EntityScheduler's executeTick of each entity. This is a bit expensive, due to ArrayDeque's size() call because it ain't a simple "get the current queue size" function, and due to the thread checks.
To avoid the hefty ArrayDeque's size() call, we check if we *really* need to execute the executeTick, by checking if currentlyExecuting is not empty or if oneTimeDelayed is not empty.
Most entities won't have any scheduled tasks, so this is a nice performance bonus. These optimizations, however, wouldn't work in a Folia environment, but because in SparklyPaper executeTick is always executed on the main thread, it ain't an issue for us (yay).
Optimizes "image in map" maps, without requiring the map to be locked, which some old map plugins may not do
This has the disadvantage that the vanilla map data will never be updated while the CraftMapRenderer is not present, but that's not a huuuge problem for us
The "distanceToSqr" call is a bit expensive, so avoiding it is pretty nice, around ~15% calls are skipped with this check
We could also check if the x,y,z coordinates are equal, but for now, let's just keep the identity check, which also helps us since Minecraft's code does reuse the original delta movement Vec3 object
The single thread per world is actually better because now we can profile each world separately. We still mimick the original thread pool by using a semaphore, this way we don't end up saturating all cores