StableRNG is not stable with SMC/PG across Julia versions

[penelopeysm]

Sampling the same model with PG across different Julia versions can yield different results even if StableRNG is used. I think it's probably not because of StableRNG, nor is it because of PG itself -- it's probably (almost certainly) because of something inside AdvancedPS. I'm not inclined to look into it, but I did just want to open an issue so that this information is out there somewhere.

[charlesknipp]

Is it not because of this:

Turing.jl/src/mcmc/particle_mcmc.jl

Lines 295 to 302 in 33e1a22

	particles = AdvancedPS.ParticleContainer(
	[
	AdvancedPS.Trace(model, vi, AdvancedPS.TracedRNG(), true) for
	_ in 1:num_particles
	],
	AdvancedPS.TracedRNG(),
	rng,
	)

According to AdvancedPS, we use Random123.Philo2x upon initializing TracedRNG. Although, this should no longer be an issue once I finally get to opening my SMC PR.