implement RPC rate limiting

[tgross]

Provide re-auth rate limiting at the Nomad servers, with the ability to observe
which users are creating the load.

---

Note to reviewers: this is a really chonky PR so I've broken it up into commits that can be reviewed individually:

• documentation, including all the metrics reference updates (~500 lines)
• the implementation of the rate limiter itself (most of the guts are here, ~550 lines)
• returning an appropriate error from the HTTP server
• getting ready to wire everything up by threading the RPCContext into all the RPC endpoints (+300/-100 line diff)
• adding the rate limit check to all the handlers (~650 lines of boilerplate)
• add some missing node secrets (which are needed for this but are currently safe without assuming mTLS)

I still need to do some more E2E testing of this PR but I wanted to get some reviewers of it so that we can pick up any remaining design or UX flaws in parallel.

cc @schmichael @jrasell @mikenomitch (ref RFC NMD-148)

[tgross]

Nothing that a lot of the approach here is going to be superseded by #13480 and follow-up PRs. I may end up just closing this PR out once all the auth and metrics work has been split out.

[tgross]

Leaving a note here that much of this code has been pulled out into pre-forwarding authentication and RPC rate metrics work planned to ship in Nomad 1.5.0. Once that work is done I'll rebase this on main and start over on any actual limiting.