Audit all reference randoms to current implementation

[wallstop]

Each random needs to be audited against the current best-implementation of the author. We need to make sure each sort that still has an upstream is appropriately attributed, and, that the code matches the reference as exactly as possible - specific constants, fallbacks, etc. The goal here is honest parity.

refs #747 (but for randoms)

[wallstop]

As part of this:

• We need drift detection (scheduled CI + issue opening)
• We need automated "compare our AST to their AST" style checks, to ensure correctness