Random assertions can fail randomly (despite seeding)

[penelopeysm]

e.g. this block in the VI tutorial:

docs/tutorials/09-variational-inference/index.qmd

Lines 154 to 162 in b53df24

	```{julia}
	#| echo: false
	let
	v, m = (mean(rand(q, 2000); dims=2)...,)
	# On Turing version 0.14, this atol could be 0.01.
	@assert isapprox(v, 1.022; atol=0.1) "Mean of s (VI posterior, 1000 samples): $v"
	@assert isapprox(m, -0.027; atol=0.03) "Mean of m (VI posterior, 1000 samples): $m"
	end
	```

Somehow, these assertions failed in one CI run: https://github.com/TuringLang/docs/actions/runs/11281015562/job/31375300973

Here's another failure of the same code block: https://github.com/TuringLang/docs/actions/runs/11281201489/job/31376048754

Here's the Gaussian mixture model tutorial failing: https://github.com/TuringLang/docs/actions/runs/11283217753/job/31382221263

Not sure how this happened as the random seed was already set at the top of the file. It seems that some function isn't respecting it.

[penelopeysm]

Turns out Random.seed! only affects calls within the same code cell. Unsure why.

[penelopeysm]

Reported upstream: PumasAI/QuartoNotebookRunner.jl#192

[penelopeysm]

Upstream issue fixed, so the remaining task is to re-enable asserts and check that they don't fail.

[penelopeysm]

Closed by #539