core: make the Rng a by-value parameter of Distribution::sample

[joboet]

Given that &mut R now implements Rng if R does (#159435), I think it makes more sense for Distribution::sample to take the Rng by-value instead of by-ref. For stateless zero-sized types like SystemRng this is more efficient and convenient, and for everything else creating a reference still works just the same.

[rustbot]

r? @jhpratt

rustbot has assigned @jhpratt.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: libs
• libs expanded to 12 candidates
• Random selection from 6 candidates

[tbu-]

For stateless zero-sized types like SystemRng this is more efficient and convenient, and for everything else creating a reference still works just the same.

Note that this PR can result in the exact thing it wants to prevent — more complex implementations of Distribution (that need to sample the RNG more than once), will need to create a reference. Chain that through multiple indirections (e.g. a struct containing a struct containing a struct) and you suddenly have &mut &mut &mut SystemRng. Before this PR, you still had &mut SystemRng, strictly better.

[joboet]

That's a very good point. I'll nominate for libs-api to make sure using by-ref is an explicit decision.

@rustbot label +I-libs-api-nominated