we want it to be included in receive timings, but not handler timings. Handler timings should reflect the time spent in the actual session PacketHandler, not in the event.
In refactors during PM4, I stripped out packet buffer caching, as it was problematic when events alter packets in undetectable ways.
However, I never cleaned this part of the code up properly after enabling DataPacketSendEvent to include multiple packets and multiple targets, so we were still individually encoding the packet(s) for every single session if the sum total of the sizes was below 256 bytes.
This change encodes packets once in the StandardPacketBroadcaster and retains their buffers to post to the session's send buffer directly if the resulting batch is below compression threshold.
This code is still not optimal (see ##5589), but fixing this brings broadcasting performance back to PM3 levels, without any of PM3's problems.
we don't translate raw string parameters anywhere else these days, so there's no reason to do so here either. The parameters array is already reduced to string[] by this point anyway.
the theoretical limit for transactions in this case is 64x9 (inputs) + 64x9 (output on crafting grid) + 64 (outputs to main slot) + 64 CraftingEventPackets = 1280.
This is an extreme case which assumes that a recipe could generate up to 64x10 (640) output items per iteration, filling every slot of the output grid, which should never occur in any reasonable circumstances.
this allows localizing disconnection screens (at least, once #4512 has been addressed) and the disconnect reasons shown on the console.
We already had disconnect messages implicitly localized in a few places, so this is just formalizing it.
This does break BC with any code that previously passed translation keys as the disconnect screen message, because they'll no longer be translated (only Translatables will be translatated now).
