Should CI environments infer `ColorChoice::Always`?

Context: I was trying to figure out why CI environments (i.e. CI=true and similar) weren't being colorized by default when using anstream to plumb output containing ANSI colors. I ended up on this choice implementation:

https://github.com/rust-cli/anstyle/blob/cbad464298a5b2b844dc9055ddcd7e4abde9d865/crates/anstream/src/auto.rs#L197-L223

If my read of that helper is right, the current logic only considers anstyle_query::is_ci() after raw.is_terminal() is satisfied. In practice that means that ColorChoice::Always branch won't be taken for CI environments, since (AFAIK) most CI environments don't actually provide a terminal (they support ANSI color codes, but they won't provide a true pty that would pass is is_terminal() check).

Given that, does it makes sense to split that check out, so that is_ci() becomes a sufficient condition for ColorChoice::Always? My intuition is "yes" (since the CI check seems to be a no-op as is), but there's also some nuance given that CI environments aren't real terminals and might only support a small subset of the escapes normally accepted by terminals (things like OSC codes in particular).

So, I don't have a super clear sense here, but I figured I'd open this in case others have a clearer picture than I do 😅. If this change makes sense, I'd be happy to send a PR for it.

Start by reading crates/anstream/src/auto.rs at the linked lines and trace how raw.is_terminal() and anstyle_query::is_ci() determine the color choice. Resolve whether CI alone should select ColorChoice::Always, including the noted ANSI and OSC-code concerns, and verify the resulting behavior in CI and non-terminal environments.