sql: embed zone configs within descriptors objects

[irfansharif]

Part of #66348. This will let us deprecate system.zones, and later provide the optimizer the cache of zone configs it needs for locality-aware planning. We'll also use embedded zone config to generate span configs for KV (#66391). We'll need to be careful around ensuring compatibility with existing backups here.

For the migration, we can first introduce a cluster version v21.1-A, which when rolled into, will disallow setting new zone configs (we'll later re-allow it after v21.1-B). We'll then recurse through all descriptors, and for each one copy over the corresponding zone config from system.zones into it.

Jira issue: CRDB-8024

[blathers-crl]

Hi @irfansharif, please add a C-ategory label to your issue. Check out the label system docs.

While you're here, please consider adding an A- label to help keep our repository tidy.

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

[ajstorm]

@irfansharif curious why this issue was labeled MT. Can you please comment?

[irfansharif]

No good reason, this came out of the multi-tenant zone configs work so likely applied the label too liberally.

[ajwerner]

This isn't something we have to do, and indeed there is no pressing reason to do it right now. It would be cleaner to associate this state more directly with the descriptors given we cache and write it in a correlated way. In reality, we'll probably more towards this when implementing transactional schema change. When we need to more generally encapsulate under a unified concept, this would be an important step.

[postamar]

I don't think we're going to do this. There exist other things-that-join-on-descriptors like comments, and we're going to be content to just have the descriptors collection mediate access to them.