Less special cases for non-rotated and non-flipped particles,
but since these are rare and the rest of the quad rendering code
has already gotten improvements there's not really any reason to keep
the specialized modes around.
The gains in GL calls were negligible and the new quad render code avoids
glVertex3f() entirely, which is a much better gain imho.
Apparently there was an interpolation scheme in place to spawn
quickly-spawning (ie more than 1 per frame) particles along a line
interpolated between the previous and current position,
but the math was incorrect and it always ended up spawning all
particles on the same spot.
Don't allocate interpolation data for RandomRotationRange when not needed
Hope this commit doesn't break some visuals; i'm not 100% sure.
Some particle files may rely on this behavior to look as intended.
This didn't seem to cause problems since apparently nothing ever depended
on the first keyframe being processed properly (ie. animationKey(),
bone commands, sounds, particle spawning, etc).
Curiously this was never an issue in the editor since the first keyframe
is hit immediately (since it keeps the time clamped to 0 when not animating),
and the first keyframe triggered ONLY when t==0 exactly.
With this change no keyframes are skipped, and during an animation loop,
the last and the first keyframe are hit together.
entity_setAnimationTime() got an extra param whether to skip keyframes
when setting the time manually (if not skipping, it'll catch up).
This commit also fixes a bug(?) that animationKey() was called with key -1
when changing animations, which was a side-effect of how the old code worked.
(I hope nothing depended on this behavior, i'm not sure but i think it's safe to fix)
also, with proper uncovering now done also for prerevealed maps,
don't set a prerevealed maps as actually revealed on load.
This applies the correct alpha now.
