Audit Codebase for Unseeded Random Number Generation

[coreyostrove]

As mentioned in recent PR #547, there are some places in the codebase where random sampling is performed by seeding either wasn't included, or had incomplete plumbing. This is a problem for cases where deterministically random number generation is needed for exact reproducibility (either in testing or manuscript preparation). There are undoubtedly other places where we've missed this scattered throughout the code base and we should audit it and correct as needed.

I'd also like to propose the following addition to our unofficial PR review guidelines (our "zen of pyGSTi"):

• If your code uses any RNG it must include the option (and requisite plumbing) to specify a user-specified seed value for deterministically random results. If deterministically random results can't be reasonably guaranteed even with seeding (e.g. even with deterministic RNG interactions with other platform specific implementation details may result in different results on different machines) and you know this to be the case you should include notes in the documentation to that effect.

[sserita]

We should also use this audit to migrate from the old NumPy RandState to the new default_rng.