Fix for double use of seed in random Particle shader variables.

[RPicster]

This PR fixes #55585

ParticleMaterial generated shaders have one aspect that I observed a couple of times when working with them.

When creating random values using rand_from_seed(alt_seed) the same value was returned and used for different calls of the function. That was because the seed is only changed when the particle is restarted.

That would result in the behaviour of creating the following random values (exemplary):

restart = true
if (restart) {
   position = rand_from_seed(alt_seed) = 1
   rotation = rand_from_seed(alt_seed) = 2
   angle = rand_from_seed(alt_seed) = 3
}
color = rand_from_seed(alt_seed) = 4
speed = rand_from_seed(alt_seed) = 5
transparency = rand_from_seed(alt_seed) = 6

but after the restart was done it would result in:

restart = false
if (restart) {
   position = rand_from_seed(alt_seed) = NOT CALLED
   rotation = rand_from_seed(alt_seed) = NOT CALLED
   angle = rand_from_seed(alt_seed) = NOT CALLED
}
color = rand_from_seed(alt_seed) = 1
speed = rand_from_seed(alt_seed) = 2
transparency = rand_from_seed(alt_seed) = 3

So certain random would appear as being "linked" - in the example above, the random color would be linked to the restart spawn position.

Before this PR:
https://user-images.githubusercontent.com/9423774/144683085-7ac472aa-d61b-479c-b29b-c5189c7ece7c.mp4

After this PR:
https://user-images.githubusercontent.com/9423774/144683228-9715fa93-f550-453d-83cd-0dc41396bcef.mp4

To fix this behavior a second seed value for all restart values was introduce.

Here is a reproduction/test project:
3.4-shader-random-problem.zip

[clayjohn]

We discussed this over RocketChat, the implementation looks great. Thanks!

[akien-mga]

Thanks!

[akien-mga]

Cherry-picked for 3.4.1.