base: introduce SQLInstanceID (and container)

[tbg]

With the advent of multi-tenancy, SQL servers can optionally be
decoupled from the (and share a) KV backend. As a consequence, the
NodeID may not be available and its use across all components that need
to function in a multi-tenant setup is deprecated. Yet, some components
require a similar unique identifier for the SQL server they're a part
of.

This new identifier is the SQLInstanceID. At the time of writing,
since we're not yet able to start true SQL tenant servers and only
exercise parts of this functionality in unit testing, it is a
randomized integer. It's possible that we'll replace it with
a per-tenant counter in production, putting it truly on par with
the NodeID, to be able to give the simple guarantee that these IDs
are never reused between different incarnations of the SQL process
for a given tenant. (Note that initially, at most one SQL tenant
server will be active per tenant at any given time).

This commit introduces an IDContainer which will aid this process by
wrapping both the NodeID and SQLInstanceID, very similar to the work
carried out for Gossip in #47972. We do not yet audit all of the call
sites to keep the initial diff container. Instead, IDContainer has a
Get() method that makes it a drop-in replacement for the previously used
NodeIDContainer. Get will be removed in a future commit while moving
all callers to either the Optional{,Err} or DeprecatedNodeID
methods.

Release note: None

[cockroach-teamcity]

This change is 

[blathers-crl]

❌ The GitHub CI (Cockroach) build has failed on 83ce9e95.

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is otan.}

[asubiotto]

are you planning on adding another commit to add the Optional calls or will that be in another PR?

Reviewed 17 of 17 files at r1.
Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @nvanbenschoten)

[tbg]

TFTR! I'll do that here, I just wanted to get a first round of review before I make code changes that could potentially rot. Will ping again when it's ready.

[nvanbenschoten]

Reviewed 17 of 17 files at r1.
Reviewable status: complete! 2 of 0 LGTMs obtained (waiting on @tbg)

---

pkg/base/node_id.go, line 88 at r1 (raw file):

type SQLInstanceID int64

// IDContainer wraps a SQLInstanceID and optionally a NodeID.

Should this be called SQLIDContainer? The key-value layer should never need one of these, so it's not a direct replacement for NodeIDContainer.

---

pkg/base/node_id.go, line 116 at r1 (raw file):

		fmt.Fprintf(&buf, "@n%s,", n)
	}
	return ""

I don't think this is what you meant.

[tbg]

bors r=asubiotto,nvanbenschoten

I packed on a lot of commits that remove uses of Get(). Will merge so that it doesn't rot over the weekend, but I'll send another PR with any additional review comments (also, I'll remove Get()).

[craig]

Build succeeded

• GitHub CI (Cockroach)