Standardize deterministic RNG in examples

[BD103]

If we are emphasising that the seeded rng is being used for deterministic purposes, then we kinda shouldn't be using StdRng in the first place. StdRng is documented as being deterministic in output, but not portable, so there is no assurance that the output will be the same on different platforms or different versions of rand. And if we are testing across different platforms, we'd then expect the output to match across all platforms. For further reference from the rand documentation, their documented expectations for portability/determinism.

Given rand minor versions might in the future change the underlying algorithm for StdRng, end users might find their tests breaking if they used StdRng in their tests after basing them on our examples. Better to instead encourage using the algorithm crates directly if they want to have that extra assurance in determinism. Currently, rand_chacha is what is the underlying algorithm for StdRng, specifically ChaCha12Rng. For game dev purposes, ChaCha8Rng would be a better fit to provide higher throughput at the expense of some entropy quality (though it will still be far better than what is probably required).

Originally posted by @Bluefinger in #12593 (comment)

I think a good PR would be to replace all usage of StdRng with ChaCha8Rng, as suggested by @Bluefinger. There are also a few examples that use hashing instead of a proper RNG function that could benefit from switching too.

[Bluefinger]

Just to add a further note to this: SmallRng has the same assurances as StdRng, as in it is deterministic in output but not portable. While StdRng currently behaves the same on all platforms (for now), SmallRng does not. SmallRng uses Xoshiro256plusplus on 64-bit platforms, but Xoshiro128plusplus on 32-bit platforms (docs reference, source code reference). So if one uses SmallRng and tries to compare output from a 64-bit desktop platform versus WASM on the web (currently wasm32), then they'd find the output to not be the same. This is because rand makes no assurance on portability for SmallRng, and since StdRng has the same lack of portability, it cannot be relied upon to be "stable" in the future. Hence the point I made above.

[TimLeach635]

I can take this one!

[BD103]

I can take this one!

Feel free to open up a PR, then!

(By the way, we don't really do issue claiming; it's a first-come-first-serve system. I appreciate your enthusiasm, though!)

[TimLeach635]

Made a PR but I'm not sure how to add labels to it (if that's the correct workflow)?

[BD103]

Closed in #12644.