Reset spiderling walk instructions on Destroy.

[warriorstar-orion]

What Does This PR Do

ACTUALLY fixes #19130.

I completely jumped the gun in #19147. It wasn't until @GDNgit pointed out the same issues with simple robots that I realized the common factor between all the GC failures wasn't any component, but the use of /walk_to.

After extensive testing I've come to the conclusion that when a /walk_to(foo, ...) call is being processed in the background by BYOND, this counts as a reference to foo for the purposes of qdel'ing it. Only a handful of objects use /walk_to. Most are simplemobs, most hostile mobs, and spiderlings, which are not a mob, but which move anyway (which is why spiderlings is happening in one PR, and the other fixes are happening later).

The fix is easy; as per the BYOND docs, "A call to a walking function aborts any previous walking function called on Ref. To halt walking, call walk(Ref,0)." (It refers to "walk", not "walk_to", in both proc's docs, but I am 99% sure that if you start with a walk_to, then halting it should also be a walk_to.

Why It's Good For The Game

GCs failures bad. If this is right (which I have incredibly high confidence in) then this means other bugs will get fixed too.

Testing

I painstakingly put these two videos together to demonstrate the before and after.

In both videos, I first show the entire diff of the running server; both cases include:

1. enabling the debugging options for GC ref finders
2. changing two trivial things about spiderling behavior: how often it calls random_skitter (from 33% of the time to all of the time) and has spiderlings grow by a constant, large amount, every process tick.

These changes are to more reliably reproduce the behavior. I think it can be pretty universally agreed that these changes have no impact on the behavior under test.

The only other difference between the two diffs is the addition of /walk_to(src, 0) in the spiderling's Destroy.

In both videos, a spiderling is spawned and grows.

In the first video, there is a period of waiting (fast-forwarded) until about 150 seconds when the GC ref finder seizes the game and starts logging the GC failure.

In the second video, there is a period of waiting (fast-forwarded) until 300 seconds when I give up waiting for the hard destroy.

Before

gc_before_2.mov

After

gc_after.mov

Changelog

🆑
fix: Fix GC failures for spiderlings
/:cl: