Optimize initialization and maploading

[out-of-phaze]

Description of changes

Performs a lot of optimizations across the codebase, some more experimental than others. I tested and got some pretty impressive results downstream on Spacefarers (loading in a new ship after roundstart took about 4 seconds instead of 16) and it seems like a lot of trouble spots in initialization are faster too.

The commit messages are descriptive, I would suggest referencing those for a description of what's included in this PR. Of particular note are these changes:

• /turf/unsimulated no longer goes through Initialize
• /obj/effect/floor_decal no longer goes through Initialize or Destroy
• update_alpha_mask now sleeps until, at minimum, the end of the tick (instead of only sleeping if the server was overloaded)
• A bunch of event registrations are now explicitly cleaned up in Destroy(), to avoid costly event cleanup code.
• Unregistering from the Moved event also unregisters from its parent. ~~(I'm unsure about this one but I couldn't see any other way to fix a bunch of hanging recursive_move event registrations.)~~ Seems to work like a charm.
• A few things were changed to respect level data bounds, which is helpful to optimize smaller away sites in large worlds.
• SSambience avoids duplicate queueing and avoids costly removal attempts when we know the turf is not in the queue.
• initial_gas = list(/decl/material/gas/oxygen = MOLES_O2STANDARD, /decl/material/gas/nitrogen = MOLES_N2STANDARD) now has a special hot-path: initial_gas = GAS_STANDARD_AIRMIX. This avoids creating a list per turf (including a unsimulated fillter turfs!), which saves memory and init time.
• Turf height appearance icons are now cached.
• Unsimulated atoms skip visualnet updates, which makes ChangeTurf() and Destroy() even faster for them.
• Extra argument handling in qdel() after force was removed. Nothing used this and it was a waste of processor time.

Why and what will this PR improve

Lots and lots of optimizations.

Authorship

One particular commit, Port InitAtom optimizations from tgstation, is adapted from some of the changes made by Mothblocks on tgstation in this commit. The rest was all done by me, mostly for the downstream Spacefarers.