Get timeseries subcommands out of experimental

See https://github.com/oxidecomputer/oxide.rs/pull/934/commits/fc95ef8a9ca93a65fec4e2406693fc979c1ea5f2

I noticed this while bumping the CLI in the v12 docs PR: the timeseries command hierarchy is a little confusing. It's all under experimental (which makes it less bad and we can probably punt for a while) but we're reproducing the system/developer distinction inside of experimental.

I think it's tolerable, but it's worth mulling over. Presumably these commands will eventually stop being experimental. In that case we have oxide timeseries (silo-scoped) and oxide system timeseries (fleet-scoped). That matches the API routes, but it's not really how we handle fleet-scoped stuff in the CLI. For example, silo stuff is all admin-oriented but it lives under oxide silo.

I'm thinking oxide timeseries is the top-level thing, and then we distinguish system vs. not inside there? It's still a little dicey, e.g., is it oxide timeseries system query vs. oxide timeseries query?

Start by reviewing the linked pull request commit and the current timeseries command hierarchy in the CLI, then compare it with the system/developer conventions described in the issue. Done means agreeing on a non-experimental command layout and updating the affected subcommands consistently; the issue does not name specific files or tests.