Accurate RTT shadow angles relative to world lights

[DESTROYGIRL]

Description

A backported feature from the Alien Swarm SDK by Saul Rennison

Current issues:

• Transitioning between two lights at near opposite angles can make the object's shadow do a very jarring sweep

• When an object stops moving it makes a very obvious flick to the high fidelity shadow at certain angles (Remote players only?)

Toolchain

• Windows MSVC VS2022

[Rainyan]

It seems that currently, the approach angle and player velocity affect the resolving of the shadow direction (or so it seems; I haven't debugged the exact reason for this).

I'm worried this would have a negative effect on gameplay, because it's unintuitive where exactly your shadow will be displayed for the opponents—even when standing in the same exact location, it's possible to have a different shadow cast outcome depending on the route you took there and your speed:

lerp.webm

(This could also be related to the sequence of closest lights travelled, with A->B resulting in a different outcome than C->B, somehow. But I'm speculating.)

---

There's also this strange issue where fidgeting around will make the shadow slowly tend towards what I assume is the neutral (FLT_MAX?) shadow position:

shadow.webm

which also would be a poor experience for the competitive meta, I fear.

[DESTROYGIRL]

I didn't consider this aspect.
We could revert to the shadow moving on it's own rather than being restricted to only moving when the player does, but it looked very silly in my opinion. But it seems that that's what is done on newer engine branches.

Maybe a system where the shadow angle is determined from the average between lights, so the movement of the shadow might not look so jarring. But then the shadows might not look accurate

[Rainyan]

As an update for this PR, I have a supplementary WIP PR open for it to address some of the shadow problems mentioned above at: DESTROYGIRL#2

I suggest we wait on this PR until DESTROYGIRL#2 is good for merge to this one, and proceed from there.